Impact of Residual Water Vapor on the Simultaneous Measurements of Trace CH 4 and N 2 O in Air with Cavity Ring-Down Spectroscopy

: Methane (CH 4 ) and nitrous oxide (N 2 O) are among the most important atmospheric greenhouse gases. A gas sensor based on a tunable 7.6 µ m continuous-wave external-cavity mode-hop-free (EC-MHF) quantum cascade laser (from 1290 to 1350 cm − 1 ) cavity ring-down spectroscopy (CRDS) technique was developed for the simultaneous detection of CH 4 and N 2 O in ambient air with water vapor (H 2 O) mostly removed via molecular sieve drying to minimize the impact of H 2 O on the simultaneous measurements. Still, due to the broad and strong absorption spectrum of H 2 O in the entire mid-infrared (mid-IR) spectral range, residual H 2 O in the dried ambient air due to incomplete drying and leakage, if not properly accounted for, could cause a signiﬁcant inﬂuence on the measurement accuracy of the simultaneous CH 4 and N 2 O detection. In this paper, the impact of residual H 2 O on the simultaneous CH 4 and N 2 O measurements were analyzed by comparing the CH 4 and N 2 O concentrations determined from the measured spectrum in the spectral range from 1311 to 1312.1 cm − 1 via simultaneous CH 4 and N 2 O measurements and that determined from the measured spectrum in the spectral range from 1311 to 1313 cm − 1 via simultaneous CH 4 , N 2 O, and H 2 O measurements. The measured dependence of CH 4 and N 2 O concentration errors on the simultaneously determined H 2 O concentration indicated that the residual H 2 O caused an under-estimation of CH 4 concentration and over-estimation of N 2 O concentration. The H 2 O induced CH 4 and N 2 O concentration errors were approximately linearly proportional to the residual H 2 O concentration. For the measurement of air ﬂowing at 3 L per min, the residual H 2 O concentration was stabilized to approximately 14 ppmv, and the corresponding H 2 O induced errors were − 1.3 ppbv for CH 4 and 3.7 ppbv for N 2 O, respectively.


Introduction
Global warming is considered to be the cause of climate change, and it has brought great consequences to human society [1]. Methane (CH 4 ) and nitrous oxide (N 2 O) are two of the most important atmospheric greenhouse gases, contributing significantly to global warming as well as climate change. Their global warming potential (GWP) for a time horizon of 100 years are 25 [2] and 298 [3] times greater than CO 2 . Due to their excellent stability and long atmospheric life periods, even small changes in N 2 O and CH 4 concentrations in air will have a long-term effect on the atmosphere [4]. Therefore, highly sensitive and precise measurements of CH 4 and N 2 O concentrations and their temporal variations in the atmosphere with ppbv (parts per billion by volume) accuracy are essential for environmental monitoring and greenhouse gas controlling [5,6]. Up to now, plenty of techniques have been developed to detect trace gases, such as non-dispersive gas sensing including non-dispersive infra-red (NDIR) absorption spectroscopy, tunable diode laser absorption spectroscopy (TDLAS), cavity ring down spectroscopy (CRDS), and photoacoustic spectroscopy (PAS) [7,8]. Mahbub et al. have recently designed rapidly pulsed near-infrared light emitting diodes (NIR LED) based nondispersive infrared (NDIR) spectroscopy at 1.65 µm for continuous remote sensing of atmospheric methane (CH 4 ), and the limits of detection (LOD) of CH 4 is 300 ppm [9]. Shao et al. have used a single DFB diode laser emitting at 2.33 µm combined with the TDLAS technique for continuously detecting atmospheric CO and CH 4 at a level of 0.73 and 36 ppbv, respectively [10]. Among these various techniques, CRDS is a direct absorption technique with a significantly improved sensitivity than the conventional direct absorption spectroscopy due to its long effective absorption path length and insensitivity to intensity fluctuations of the light source [11][12][13], which has been widely used to detect trace gases in real time with ultrahigh sensitivity and relatively low system complexity in recent years [14]. With the rapid development of mid-infrared (mid-IR) quantum cascade lasers (QCL) and QCLs utilized in CRDS to employ the strongest absorption lines for trace gas detection, the sensitivity of CRDS was further improved [15][16][17] and has achieved the required sensitivities for real-time monitoring of trace species from ppmv (parts per million by volume) down to the pptv (parts per trillion by volume) levels.  [20].
However, the presence of gases other than the target species during the measurement will seriously affect the measurement accuracy [21][22][23][24][25]. Water vapor, one of the most important interference gases in the atmosphere, has a wide absorption spectrum from microwave to far-IR, and especially in the mid-IR spectral region. The spectrum of H 2 O cited from the HITRAN database [26] is shown in Figure 1a. Additionally, the concentration of H 2 O in the atmosphere varies in a large dynamic range from nearly 0 up to 4% over temporal and spatial scales [27] and mixing ratios of CH 4 and N 2 O are significantly affected by variations of water vapor. Moreover, without drying almost no CRD signals can be experimentally detected by the CRDS experimental setup used in this paper due to the high H 2 O concentration (in 1% level) in the ambient air and the limitation of the maximum detectable absorption coefficient (~3 × 10 −5 cm −1 , corresponding ring-down time is about 1.5 µs) of the setup. According to the HITRAN database, the spectral absorptions for a typical ambient concentration of CH 4 (2 ppmv) and N 2 O (0.3 ppmv) with an H 2 O concentration of 10 ppmv and 1.4%, respectively are given in Figure 1b. Generally, H 2 O in the ambient air can be largely removed by passing the air over molecular sieves [28][29][30] and the concentration of H 2 O can be effectively decreased to even less than 1 ppmv [20]. In this case, its impact on the CH 4 and N 2 O detection becomes negligible. However, it is possible that if over-drying, the molecular sieve might also adsorb target gases and in turn affects the retrieval of the target quantity due to its non-uniformly distributed pore sizes. A more practical way may be reducing the amount of molecular sieve to minimize its influence on the target gas detection. In this case, the residual H 2 O in the levels from several ppmv up to 100 ppmv exists in the dried air due to incomplete drying. In addition, a possible leakage in the gas-handling system may also cause the increase of H 2 O concentration inside the sample cell. In this paper, a CW-CRDS experimental setup with a mode-hop-free (MHF) EC-QCL operating between 1290 to 1350 cm −1 is applied to simultaneously detect CH 4 and N 2 O in clean laboratory air. In this experimental setup, the H 2 O is dried to a concentration from several ppmv to 200 ppmv by passing the measured air through molecular sieve filling tubes. A method is developed to determine simultaneously the concentrations of CH 4 , N 2 O, and residual H 2 O by measuring the absorption spectrum in a spectral range covering the absorption lines of CH 4 , N 2 O, and H 2 O. The impact of residual H 2 O on the simultaneous CH 4 and N 2 O measurements is analyzed by comparing the results that with and without determining simultaneously the concentration of residual H 2 O is increased with time due to the slow leakage of the laboratory air into the closed ring-down cavity (sample cell) of the CRDS setup. Then, the same method is applied to analyze the impact of residual H 2 O on the simultaneous CH 4 and N 2 O measurements in flowing air.

Experiments
The apparatus used for the CRDS experiment is similar to that described in [20] and is schematically depicted in Figure 2. In brief, a tunable mid-IR external-cavity CW-MHF QCL (41074-MHF, Daylight Solutions, San Diego, California, USA) with an MHF tuning range of 1290-1350 cm −1 and power nearly 160 mW is utilized as the light source. The QCL emits a collimated laser beam with a narrow spectral linewidth (<30 MHz or 0.001 cm −1 ) which meets the requirements for high-precision spectral measurements. Moreover, the QCL is tuned via the laser controller with a step of 0.01 cm −1 (with accuracy < 0.003 cm −1 ). In order to avoid back reflections of light from other optical components affecting the laser output, an optical isolator with central wavelength 7.2 µm and isolation ratio >30 dB (FIO-5-7.2, Innpho, Verona, NJ, USA) is placed in front of the laser. Then, the light passes through an acousto-optic modulator (AOM) (I-M041, Gooch and Housego, Ilminster, UK) acting as a fast-optical switch that produces the zeroth and first-order diffraction beams. The AOM is controlled by a high-speed (with response time < 50 ns) threshold trigger circuit. The first-order diffraction beam is coupled into the stable high-finesse ring-down cavity which consists of a 50 cm long quartz-coated sample cell and two high-reflectivity planeconcave mirrors (CRD Optics) with reflectivity R > 99.98% and 1 m radius of curvature attached to two ends of the ring-down cell. Three piezoelectric transducers (PZT, Model PE-4, Thorlabs, Newton, NJ, USA) mounted on the rear cavity mirror is applied to achieve the periodic laser-cavity coupling. A triangular signal (Vpp = 5 V, frequency = 40 Hz) is applied to the PZTs to modulate the cavity length over one half of the wavelength. A periodic resonant signal (TEM 00 mode) is built-up periodically inside the cavity due to mode-matching between the laser and cavity. The QCL beam leaking from the cavity is focused by a focusing lens and subsequently detected by a TE-cooled HgCdTe infrared photovoltaic detector (PVMI-4TE-8, Vigo, Poland). When the measured light intensity reaches a preset voltage value, the threshold trigger circuit sends a trigger signal to the AOM to switch-off the first-order beam. The ring-down signal is then recorded by a data acquisition (DAQ) card (M2i.3010, Spectrum Instrumentation, Großhansdorf, Germany) and analyzed by a MATLAB program. A vacuum pump (MPC 301Z, Welch, Concord, MA, USA), a pressure gauge (LEX1, Keller, Winterthur, Switzerland), and a mass flow meter (EL-FLOW Select, Bronkhorst, The Netherlands) are used to control the pressure and gas flows inside the sample cell. Three plexiglass drying tubes filled with 3A molecular sieves (Rhawn, Shanghai, China) which can absorb molecules with a dynamic diameter less than 0.3 nm are connected to the gas inlet of the sample cell to remove the water vapor from the measured gas mixture flowing into the sample cell. It is experimentally found that only when the concentration of water vapor in the measured air is below a certain level, the CRDS signal becomes observable and measurable. Differed from that in [20], no 3A molecular sieves are put inside the sample cell for further drying in order to minimize the possible adsorptions of CH 4   Experimentally by determining the ring-down time τ from the measured ring-down signal, the wavelength-dependent absorption coefficient α(λ) of the target gas can be obtained by where c is the speed of light, λ is the wavelength, and τ 0 is the ring-down time in the empty sample cell (in our case, the "empty" sample cell is filled with high-purity nitrogen which has negligible absorption in the measurement spectral range). The absolute concentration of the target gas can be determined by comparing the measured absorption spectrum to the HITRAN database. For the simultaneous determination of multiple gases, the measured absorption spectrum can be expressed as where c i is the concentration of gas i (i = 1, 2, . . . , N) in the gas mixture, and α i (λ) is the corresponding absorption spectrum, which can be obtained from the HITRAN database. Once the number N is determined, the concentrations of the multiple target gases can be determined by multi-parameter fitting the measured absorption spectrum to Equation (2). The absorption of interference gases can be treated as the background term which is set as a free parameter in the multi-parameter fitting. From Equation (2), to minimize the influence of background absorption on the simultaneous determination of multiple target gases, the level of background absorption should be as low as possible compared to the absorption caused by the target gases. In this case, the selection of wavelength range for the absorption spectrum measurement of multiple target gases is essential. The gas mixture used in the experiment is the ambient air collected from the laboratory room with H 2 O partially removed by molecular sieves. The relative moisture level in the laboratory room fluctuates between 50% and 60%, which represents a H 2 O concentration range of 1.4-1.7%. The ambient air enters the sample cell (the ring-down cavity) through a vacuum valve (MPC 301Z, Welch). All the experimental data are recorded in the clean-room laboratory with approximately 296 K temperature and 950 mbar pressure (or 1 atmospheric pressure), respectively. For the simultaneous determination of CH 4 and N 2 O, N is set to 2 in Equation (2), and the impact of H 2 O is included in the background term. The concentration of H 2 O can also be simultaneously determined by setting N = 3 in Equation (2) and H 2 O being treated as an additional target gas when performing the multi-parameter fitting.

Spectral Range Selection
To achieve simultaneous measurements of CH 4

Impact of Residual Water Vapor on CH4 and N2O Measurements
Firstly, the gas inlet and outlet valves connecting to the sample cell are open and laboratory air is fed into the sample cell by the vacuum pump at a flow rate of 3 L per min for 5 min. The H2O in the sample cell is partially dried due to the molecular sieves inside the plexiglass drying tubes. Then, the gas inlet and outlet valves are adjusted to make the air pressure in the sample cell approximately equal to the outside air pressure (about 950 mbar), and then the valves are closed to form a closed sample cell filling with ambient air with H2O partially removed. The experiment is performed to continuously measure the laboratory air for about 7 h from 13:30 to 20:30 on 30 July 2019. The QCL is tuned from 1311 to 1313 cm −1 in 20 min and the measurement is repeated every 50 min (repeated 8 times in 7 h). The concentrations of CH4, N2O, and H2O are then determined via fitting the measured spectra to Equation (2) with N = 3. The measured H2O concentration is presented in Figure 4. An approximately linear relationship between the H2O concentration and the measurement time is observed. This linear increase of H2O concentration inside the sample cell with time is due to leakage of the sample cell and the gas system. From the linear dependence, the leak rate of H2O can be determined to be approximately 27 ppmv/h. On the other hand, the intercept of 15 ppmv indicates the drying capacity of the molecular sieves at the flow rate of 3 L per min.  On the other hand, the CH 4 and N 2 O concentrations can also be determined by employing the spectrum measured in the spectral range from 1311 to 1312.1 cm −1 . In this case, the concentration of H 2 O is set to zero and its influence on the CH 4 and N 2 O measurements is included in the background term in Equation (2). In the spectral range from 1311 to 1312.1 cm −1 , there are two N 2 O absorption lines and one CH 4 absorption line, which are used to determine the CH 4  In Figure 6, the first and eighth measured and fitted spectra from 1311 to 1313 cm −1 and the corresponding residuals are shown. In the eighth measurement, the peak absorption of H 2 O is significantly higher than that of N 2 O and CH 4 and many data points around the H 2 O peaks become undetectable due to the reduced signal-to-noise ratio (SNR) of the ring-down signals. It is clearly indicated that the fitting residual of N 2 O is larger than CH 4 , since the second absorption peak of N 2 O is closer to the absorption peak of H 2 O and is more affected by the changes of H 2 O concentration.

CH 4 and N 2 O Measurements of Flowing Air
In practical applications, the CH 4 and N 2 O concentrations in the ambient air need to be monitored continuously. In this case, the measured air is flowing through the sample cell. Therefore, the simultaneous measurement of CH 4 Figure 7. Obviously, the H 2 O concentration initially shows a downward trend in the first 3 h, and then gradually becomes stabilized to around 13-15 ppmv, which is consistent with the value (15 ppmv) obtained in Section 3.2 showing the drying capacity of the molecular sieves at the flow rate of 3 L per min. Due to the impact of residual H 2 O, the CH 4 and N 2 O concentrations are under-estimated by approximately −1.3 ppbv and over-estimated by approximately 3.7 ppbv, respectively. These H 2 O induced measurement errors are somewhat smaller than that estimated with the stable air in the closed sample cell (Section 3.2). From the measurement results with the flowing air, the residual H 2 O induced errors can be corrected once the H 2 O concentration becomes constant. Therefore, in real applications, once the H 2 O concentration becomes stabilized, there is no need to measure the residual H 2 O concentration repeatedly to reduce the measurement time (the time for one measurement is reduced by nearly half from 20 min to nearly 10 min in our case). From the results presented in Figures 5 and 7, the fluctuation of the CH 4 concentration during the whole measurement period is larger than that of N 2 O, which is in agreement with our previous measurement results [20].

Discussion
In this paper, the influence of residual H 2 O on the simultaneous measurement of CH 4 and N 2 O trace greenhouse gases in ambient air with tunable mid-IR CRDS coupled with EC-QCL in the spectral range from 1311 to 1313 cm −1 has been investigated by treating the residual H 2 O either as a background gas or as a target gas, whose concentration is being determined simultaneously with the CH 4