Detection of HO2 in an atmospheric pressure plasma jet using optical feedback cavity-enhanced absorption spectroscopy

Cold non-equilibrium atmospheric pressure plasma jets are increasingly applied in material processing and plasma medicine. However, their small dimensions make diagnosing the fluxes of generated species a challenge. Here we report on the detection of the hydroperoxyl radical, HO2, in the effluent of a plasma jet by the use of optical feedback cavity-enhanced absorption spectroscopy. The spectrometer has a minimum detectable absorption coefficient α min of 2.25 × 10 − 10 cm−1 with a 100 second acquisition, equivalent to 5.5 × 10 12 cm − 3 of HO2 (under ideal conditions). Concentrations in the range of (3.1–7.8) × 1013 cm−3 were inferred in the 4 mm wide effluent of the plasma jet.


Introduction
The hydroperoxyl radical, HO 2 , plays an important role in atmospheric chemistry, as it is a key radical in the OH-induced oxidation cycle which controls virtually all of the daytime oxidative chemistry in the lower atmosphere and is therefore responsible for the transformation of primary emissions into secondary pollutants [1]. Furthermore, HO 2 is prevalent in hydrocarbon combustion [2][3][4] and in dielectric barrier discharges utilized for the purification of water [5] or the removal of volatile organic compounds in air [6].
The HO 2 radical is also considered to be an intrinsic part of the chemical reaction network in cold nonequilibrium atmospheric pressure plasmas, due to its involvement in the generation pathways of the hydroxy radical, OH, one of the most studied reactive species in plasma jets [7][8][9][10][11]. These plasma jets have gained strong attention over the past decade due to their use in the field of plasma medicine [12], where the high reactivity at low gas temperature is crucial for interaction with sensitive biological systems [11]. This makes them ideally suited for treatment of human tissue in plasma-assisted wound healing, cancer treatment [13], or sterilization [14]. The OH molecule is generated via pathways which include HO 2 , which itself is quickly quenched by air derived species [10], and therefore HO 2 is mostly confined to the localized afterglow of the plasma jet. Due to their small size and high density gradients in space and time, these jets, however, are difficult to diagnose quantitatively. Absorption spectroscopy has become a popular method for characterizing the fluxes of species generated by these plasma sources [15]. To achieve space-resolved densities from the line-of-sight integrated densities, Abel inversion is most commonly used, which leads to a local density as a function of distance from the centre [16]. However, the generally small geometry of the effluent of a plasma jet (in μm to cm range) severely limits the sensitivity of the method, especially for highly reactive transient species.
To overcome the difficulty of small absorption lengths, cavity-enhanced spectroscopy is a promising method for localized measurements of species with low abundances in plasma jets. A wide class of cavity-enhanced spectroscopy techniques have been developed such as cavity ring-down spectroscopy (CRDS), off-axis cavityenhanced absorption spectroscopy (CEAS), optical feedback cavity-enhanced absorption spectroscopy (OF-CEAS) and noise-immune cavity-enhanced optical heterodyne molecular spectroscopy (NICE-OHMS) [17,18]. Of these variants, CRDS has been the method of choice for the detection of HO 2 utilizing absorption bands in the near-infrared (NIR) range [3,[19][20][21], although its detection using NICE-OHMS was also reported [22]. Similarly, CRDS has been employed for the detection of species in plasmas. An overview of the existing applications of CRDS to characterize various types of atmospheric pressure plasma jets can be found in [23]. In this paper, we utilize the related technique of OF-CEAS in which the laser is scanned over the spectral region of interest and locks, through optical feedback, to successive cavity resonances. Consequently, the optical cavity is used to both greatly increase the optical pathlength through the sample and enhance the measured signal intensity, resulting in very low detection limits being achievable [18,24]. Initially, NIR diode lasers were used as sources [25,26], but recently OF-CEAS instruments based on quantum and interband cascade lasers have also been reported [27][28][29][30][31].
Liu et al have measured the concentration of HO 2 in a dielectric barrier discharge plasma using NIR continuous wave (cw) CRDS and investigated the effect of water on the concentration of HO 2 in a CH 2 O/O 2 / H 2 O/N 2 gas mixture [20]. These studies, however, were conducted at a pressure of 30 Torr and with a plasma sample length of 10 cm resulting in a detection limit of´-1 10 cm 11 3 . In this article, we present first measurements of HO 2 in the effluent of a mm size atmospheric plasma jet, the so-called kINPen-Sci, employing OF-CEAS using a diode laser at 1506.43 nm as the light source to probe HO 2 transitions in the the first vibrational overtone of the O-H stretch, the n 2 1 band (centred at 6649 cm −1 ). We begin our discussion by presenting the construction and performance of our OF-CEAS spectrometer. Figure 1 is a schematic representation of the OF-CEAS spectrometer. The beam from a cw distributed feedback laser diode (LD, l = 1506.4 nm, = P 32 mW, NTT Electronics NLK1S5D1TA) housed in a temperaturestabilized mount (Thorlabs TCLDM9) is collimated by a = f 2.5 mm lens and directed via protected silver mirrors (Thorlabs PF10-03-P01) to a delay line (DL), a half-wave plate (HWP, Thorlabs WPH05M-1310), a polarizing beam splitter (PBS, Thorlabs PBS104), and into the V-shaped cavity. Current and temperature of the laser are controlled by two modules (Thorlabs LDC8002 and TEC8020) housed in a chassis (Thorlabs PRO8000). The polarization axis of the linearly polarized beam can be rotated with the HWP (mounted in a rotation mount, Thorlabs RSP1), so that the intensity ratio between the transmitted and reflected beams after the PBS can be adjusted. The intensity of the beam reflected by the PBS is measured by an InGaAs photodiode (PD2, Thorlabs DET410). After the PBS, a mode-matching lens with = f 500 mm (Thorlabs LA1908-C) focuses the beam to the waist size and position required to excite the fundamental transverse cavity mode. The last mirror before the cavity (MPZ) is mounted on a piezo-electric transducer. The piezo-mounted mirror and the DL allow for fine and coarse adjustment of the distance between the laser diode and the folding mirror (FM) of the cavity. The optical cavity consists of three highly reflective dielectric mirrors (Layertec) with an effective reflectivity > 0.9999, arranged in a V shape with an angle of 5.5°and an arm length of 85 cm. The external reflection off the FM is collected by a beam dump (BD). A Vidicon camera (VD, Electrophysics MicronViewer 7290A) placed at one of the output ports allows monitoring of the beam profile, while an InGaAs photodiode (PD1, Thorlabs DET10C/M) measures the transmitted intensity through the cavity. A transimpedance amplifier (Femto DLPCA-200) with the gain typically set to 10 6 V A −1 amplifies the photocurrent from PD1 prior to digitization by a USB data acquisition module (National Instruments NI-USB 6356). The signal from PD2 does not require amplification. The spectrometer is controlled by a custom LabVIEW (National Instruments) program running on a standard PC. The beam in one of the arms of the V-shaped cavity intercepts the effluent from a kINPen plasma jet device (PJ). The cavity is enclosed by an aluminium box with inner dimensions of´93 23 15 cm (width×depth×height) and an acrylic glass lid.

Laser spectrometer
The theoretical framework of optical feedback-based spectrometers and their advantages are described in detail elsewhere (see for example [18], chapter 5), so the working principle will only be outlined briefly here. Optical feedback techniques exploit the very strong response of a diode laser to radiation intentionally allowed to return to it. Under appropriate conditions, when radiation returns from a high-finesse cavity to the laser source, optical feedback can: (a) significantly reduce the emission linewidth of the laser, e.g., from MHz to below 1 kHz, thus increasing the throughput and the coupling efficiency of the laser with the cavity; (b) lock the emission frequency of the laser to a cavity resonance. As the laser frequency is scanned (by ramping its current or temperature), the laser locks to successive cavity modes, ideally with little or no time period during which the laser is free-running ('unlocked'). This behaviour is illustrated in figure 2(a), where the black trace shows the raw cavity transmission spectrum. A weak water transition is present in the measured frequency interval.
One of the parameters that plays an important role in the context of optical feedback is the exact distance between the laser facet and the entry mirror of the cavity (the folding mirror in a V-shaped cavity). In order for optical feedback to successfully lock the laser emission frequency to any cavity mode, the distance D between the laser facet and the entry mirror must be equal to = figure 1), L 2 is the arm length FM-M2, and  q 0 is an integer [18]. In our case, q=0 and = = D L 85 cm 2 . It can be shown that the shape of the modes (see black trace in figure 2(a)) depends on this condition being fulfilled. When that is the case, the modes have a roughly symmetric appearance, but when ¹ D  L 2 they become more and more asymmetric, depending on the sign of -D L 2 [32]. This asymmetry is used here as an error signal to maintain the distance D at its correct value: an average asymmetry value is computed for the modes in each scan, and this value is then used to adjust the voltage applied on the piezo (MPZ in figure 1).

Extraction of quantitative data
The transmission maximum of each mode can be evaluated from the transfer function of a V-shaped cavity at the kth resonance [18]: where T and R are the mirror (intensity) transmission and reflection coefficients, L is the sum of the two arm lengths of the cavity, and a a = -- is termed the equivalent absorption coefficient of the cavity (assuming the mirror's contribution to the cavity losses to be distributed over the length of the cavity). Equation (1) can be solved with respect to ã k to yield The maximum H k of each mode is extracted experimentally from the raw cavity transmission (red circles in figure 2(a)). The second equality in equation (2) exploits the relationship between the ring-down time τ and the  where S is the line strength (or integrated absorption cross-section) and A is the integrated absorption coefficient. For the water absorption line in figure 2 we obtain =´-N 4.8 10 cm 13 3 . The measured full width at half maximum (FWHM) of  586 3 MHz is close to the theoretical value of 580 MHz at 298 K. As the frequency axis in figure 2 is determined via the free spectral range of the cavity, which in turn is computed from the measurement of the two arm lengths, the small discrepancy in width could be explained by an error in the cavity length measurement.

Plasma jet
The non-thermal plasma used to generate the HO 2 is an argon atmospheric pressure plasma jet operated at a radio frequency around 1 MHz ( figure 3). The so-called kINPen-Sci has a pin type powered electrode centred in a dielectric tubing with a circular grounded electrode outside the dielectric [33]. The plasma jet is operated at about 1W dissipated plasma power. The energy is quickly transferred to metastable states of argon or nitrogen [10,34]. The plasma jet is equipped with a gas curtain device which allows control of the atmosphere surrounding its active afterglow [35]. Changing the composition of the curtain gas strongly influences the reactive component composition generated by the plasma. For instance, humidity variation controls the output of H 2 O 2 and OH [36]. The plasma jet is introduced into the cavity via a side opening. An xy-stage allowed a precise positioning of the effluent of the plasma jet to intersect the laser beam. The laser beam of 1 mm diameter passed through the plasma effluent at 11 mm distance from the plasma jet's nozzle as this is the typical distance between plasma jet and treated surfaces. The feed gas argon has a flow rate of 3 standard litre per minute (slm) through the plasma jet. It is humidified by passing a small fraction (4%-10%) through a water bubbler held at room temperature. The feed gas humidity is measured with a chilled mirror dew point hygrometer (EdgeTech DewMaster) according to the procedure described previously [36]. The gas curtain is operated with a mixture of 20% oxygen (purity 2.6) to 80% nitrogen (purity 4.8) at a flow rate of 5 slm. All gas flows are controlled via electronic mass flow controllers (MKS Instruments). , respectively, the waves propagating in the two arms of a fibre interferometer [38]. The beam was coupled into the fibre via a fibre collimation package (FCP) placed instead of the reference detector (PD2 in figure 1). After recombination of the two arms of the fibre interferometer, a fast photodiode (PD, Hamamatsu G8372-01), transimpedance amplifier (AMP, FEMTO DHPCA-100), and a spectrum analyzer (SA, HP 8560E) measure the beat note signal.
Assuming white phase noise and t t d c  , where t c is the coherence time of the beam, the power spectral density (PSD) of the beat note signal has the same (Lorentzian) shape as the laser emission spectrum, but with a width twice as large [38]. The free-running laser exhibits a linewidth of 1.3 MHz, which reduces to 3.4 kHz when the laser is locked to the cavity ( figure 5(a)). The condition t t d c  is fulfilled in the free-running case (t m < 1 s c ), when the beam is blocked between the mode-matching lens and the cavity, but not in the case with optical feedback. For this reason, the linewidth of 3.4 kHz quoted above is not accurate. A better value could be obtained by modelling the PSD of the beat note signal, for example as in equation (17) in [38], but this is beyond the scope of this article. Nevertheless, the effect of optical feedback on the laser emission spectrum is evident from figure 5(a), at least qualitatively. The reduced linewidth dramatically increases the coupling efficiency between laser and high-finesse cavity and is one of the advantages of OF-CEAS.  Allowing 7.8 Torr of air into the cavity reduces its finesse due to absorption of the laser light by the water continuum. The increase in cavity losses is measurable as a decrease in ring-down time or, equivalently, as a broadening of the cavity resonance width and, through optical feedback, of the laser linewidth. This phenomenon is clearly visible in figure 5(b), where the PSD of the beat note signal is shown under conditions of vacuum ( < p 0.1 Torr) and with 7.8 Torr of air. Again, as the condition t t d c  is not fulfilled, the quoted linewidths are not accurate. In figure 5(b) we merely aim to show the qualitative increase in laser linewidth due to larger cavity losses.

Sensitivity
A typical scan covers about 270 modes or 24 GHz ( -0.8 cm 1 ) and takes 2 s to complete. There is a short dead time of about 0.25 s after each scan to compute the error signal needed for the laser-to-cavity distance adjustment. The acquisition rate is about -0.44 s 1 or 800 spectra in 30 min. By averaging multiple spectra together, noise can be reduced at the expense of a longer measurement time. We define the baseline noise as the standard deviation (SD) of the absorption coefficient α measured over a region free of absorption. Figure 6 shows the SD of α as a function of the number of spectra averaged together. Averaging more than about 50 scans does not significantly reduce the noise but simply increases the duration of the measurement. With 50 averaged spectra, the entire series can be acquired in 100 s, and the SD of α is a =´--2.25 10 cm min 10 1 . We can then estimate the limit of detection (LOD) for HO 2 by assuming the smallest absorption we can reliably detect to be a =´- 1 ); the factor p 2 arises from the lineshape, which we assume to be Lorentzian. The value of N min in equation (5) assumes that absorption is homogeneous along the full length (1.7 m) of the cavity.
In order to determine absolute concentrations of the HO 2 radicals in the plasma jet, an assumption has to be made on the pathlength over which the radicals absorb. From quantitative schlieren diagnostics on the plasma jet an absorption length of 4 mm for the plasma effluent at 11 mm distance from the plasma jet's nozzle was obtained [16]. In our calculations we also assume a homogeneous concentration distribution of the radicals over the absorption pathlength. If the absorption is confined to a 4 mm region, the LOD increases by a factor = 1.7 m 4 mm 425 to =´-N 5.5 10 cm min 12 3 . This LOD is valid for a zero flow in the cavity. When the kINPen device is active, however, the gas flow and the discharge cause very strong fluctuations in the cavity transmission, and the LOD worsens. The turbulent gas flow destroys the coherence of the optical system and makes the optical power build-up within the cavity more sporadic [40]. A very large number of spectra must then be acquired, most of which have to be discarded because of excessive noise or loss of frequency locking. In the following measurements, we typically acquired 800 spectra (30 min), of which only 200-300 were usable and averaged together.

HO 2 spectroscopy and relationship between H 2 O and HO 2
The absorption feature of HO 2 selected for the investigation in this work consists primarily of two transitions assigned by DeSain et al [41] as the ( ) P 6 q 0 transition at -6638.11 cm 1 , consisting of two transitions due to spinrotation splitting, and the overlapping ( ) N Q q 3 transitions with = -N 4 9 at -6638.20 cm 1 , consisting each time of two transitions due to spin-rotation splitting. The Doppler-limited peak absorption cross-section (at zero pressure) for the latter transition was reported at 296 K to be´- 1 . In figure 7, the measured spectra for four different feed gas humidities without and with the plasma switched on are shown. The water concentration was determined by a dew point measurement and is rounded to 10 s of ppm. We did not deduce the water concentration from the measured absorptions since these are due not only to the water in the plasma jet, but also to water absorbing along the full length (1.7 m) of the cavity (the optical cavity is open to laboratory air). HO 2 , on the other hand, being quickly quenched by air derived species, does not diffuse within the cell but is localized in the jet region. With the plasma switched off, the measured absorption spectrum is due to three water transitions at   'plasma on' spectrum is much noisier due to the plasma disturbing the stability of the cavity. A composite fit to the absorption features for a feed gas humidity of 4000 ppm is shown in figure 8. As the measurements were performed at atmospheric pressure, to obtain the area under the HO 2 absorption feature (the integrated absorption in equation (4)), the absorption spectra were fitted to five Lorentzian profiles. The absorption feature for HO 2 was fitted by a double Lorentzian with the relative centre frequencies and areas fixed and the same Lorentzian linewidth for both. The water transitions were also included to account for their overlap with HO 2 . The HO 2 concentrations as the feed gas humidity was changed from 1000 to 4000 ppm were in the range of  ´-( )-( ) 3.1 0.3 7.8 1.0 10 cm 13 3 . The main uncertainty in the determination of these concentrations is the absorption pathlength of 4 mm for the plasma effluent at 11 mm distance from the plasma jet's nozzle. The number densities are in broad agreement with predictions from modelling of these plasmas [10]. We note, however, that such cavity-based measurements provide a new way of testing and improving our modelling of these complex environments.

Conclusion
In summary, we have presented an OF-CEAS experiment and demonstrated its first application to the detection of HO 2 in the effluent of a plasma jet. We demonstrated a sensitivity of´--2.25 10 cm 10 1 with a 100 seconds acquisition time under ideal conditions, which translates into a sensitivity of´-5.5 10 cm 12 3 of HO 2 assuming an absorption length of 4 mm. In the plasma jet, HO 2 concentrations in the range of (3.1-7.8) × 10 13 cm −3 were detected. The detection levels indicate that such a spectrometer will find broad application in future studies of the chemical network in the effluents of plasma jets.