4.1 Calibration

Washenfelder et al. (2008) described a procedure for retrieving the mirror reflectivity curve by taking advantage of a different Rayleigh scattering contribution to the cavity losses while the measuring cell was filled with different bulk gases (e.g. helium versus air or nitrogen). The Rayleigh's empirical cross sections used by Min et al. (2016) and the theoretical ones used by Thalman and Volkamer (2010) were found to disagree, leading this work to propose an easier approach consisting of using trace gases at known concentrations (in this case NO₂ produced by the calibrator FlexStream™ Gas Standards Generator from KIN-TEK Analytical, Inc. and CHOCHO produced by evaporating a glyoxal solution 40 % in water from Sigma-Aldrich, since the O₃ spectrum is less structured and IO is highly reactive) and their literature cross sections (Vandaele et al., 1998, for NO₂ and Volkamer et al., 2005, for CHOCHO) for retrieving the wavelength-dependent reflectivity curve. Such an approach to calculate mirror reflectivity has been proposed before (Venables et al., 2006) and has been used by previous studies (Duan et al., 2018). The shape of the reflectivity curve is adjusted to best match the convoluted literature NO₂ spectrum at the concentration given by the KIN-TEK calibrator (for more information, see Sect. S2). Figure 2a shows the resulting reflectivity curve and the transmitted light through the cavity and the optical filter. The maximum reflectivity achieved with both mirrors given by the calibration procedure is 99.9905 %, leading to an effective optical path length of ∼4.4 km and a cavity finesse $\left(F=\frac{\mathit{\pi }\sqrt{R}}{(\mathrm{1}-R)}\right)$ of ∼33 100. While the shape of the mirror reflectivity curve is determined once and for all, its offset is slightly adjusted after each mirrors' cleaning, by flushing in the cavity a known concentration of NO₂. The spectral emission of the LED centred at 445 nm is well suited also for the detection of IO, CHOCHO and O₃, which are other key species in atmospheric chemistry. For the field measurements of NO₂ and NO_x, two twin instruments named IBBCEAS-NO₂ and IBBCEAS-NO_x are deployed, with the later equipped with an ozone generator on the gas inlet line. At this wavelength region, water vapour also absorbs and is accounted for in the spectral-fit analysis. However, the absorption of oxygen dimer is not required in the fit routine, since the absorption feature will be present in the reference (I₀) as well as in the absorption (I) spectra. In Fig. 2b simultaneous detection of NO₂, IO and O₃ is reported. Ozone, at 28.9 ppm, is produced by water electrolysis as described in Sect. 4.4; 191.8 ppb of NO₂ is provided by a permeation tube, and 425.6 ppt of IO is generated by photochemical reaction of sublimated iodine crystals and ozone in the presence of radiation inside the cavity. For this spectrum, the light transmitted is integrated for 350 ms on the CCD and averaged over 1000 spectra, yielding to a 1σ standard deviation of the residuals of $\mathrm{4}\times {\mathrm{10}}^{-\mathrm{8}}$ cm⁻¹ (Fig. 2b bottom). In Fig. 2c simultaneous detection of NO₂, CHOCHO and H₂O is reported. CHOCHO at 4.3 ppb is produced by evaporating a glyoxal solution (40 % in water; Sigma-Aldrich) at the sample gas inlet of the instrument; NO₂ at 1.40 ppb and H₂O at 0.54 % are ambient concentrations observed in the laboratory during the experiment. For this spectrum, the light transmitted is integrated for 250 ms on the CCD and averaged over 1000 spectra, yielding to a 1σ standard deviation of the residuals of $\mathrm{5}\times {\mathrm{10}}^{-\mathrm{9}}$ cm⁻¹ (Fig. 2c bottom). The top of Fig. 2b and c shows the experimental spectra (black traces), the fit result (red traces) and contributions from each species (middle) which are included in the spectral fit. The concentrations of the species are retrieved with respect to the literature cross sections using the calibrated reflectivity curve discussed above.

Figure 2(a) The mirror reflectivity curve (red) in comparison with the spectrum of the LED light transmitted by the cavity and the optical filter for a single acquisition of 250 ms. (b) In black, an example of an experimental spectrum of NO₂, IO and O₃ at concentrations of 191.8 ppb, 425.6 ppt and 28.9 ppm, respectively; in red, the multi-species spectral fit; and in blue, orange and grey the absorptions of the different species. At the bottom (in black) the residual of the experimental fit with a 1σ standard deviation of $\mathrm{4}\times {\mathrm{10}}^{-\mathrm{8}}$ cm⁻¹ after 1000 averages. (c) In black, an example of an experimental spectrum of NO₂, CHOCHO and H₂O at concentrations of 1.40 ppb, 4.3 ppb and 0.54 %, respectively; in red, the multi-species spectral fit; and in blue, orange and green the absorptions of the different species. At the bottom (in black) the residual of the experimental fit with a 1σ standard deviation of $\mathrm{5}\times {\mathrm{10}}^{-\mathrm{9}}$ cm⁻¹ after 1000 averages.

Download

4.2 Instrumental calibration and intercomparison

A calibrator (FlexStream™ Gas Standards Generator, KIN-TEK Analytical, Inc.) was used to produce a stable NO₂ source. The sample was produced using a permeation tube of NO₂ (KIN-TEK ELSRT2W) calibrated at an emission rate of 115 ng min⁻¹ at 40 ^∘C loaded into the calibrator. This type of calibrator is ideally suited for creating trace concentration mixtures (from ppt to ppm). The instruments were calibrated against the calibrator delivering NO₂ at 49.59±0.17 ppb. To confirm the calibration process as well as the stability of the instrument within a greater range of concentrations, two intercomparisons of the IBBCEAS with two different CLD instruments (Thermo Fisher™ 42i NO_x analyser and Thermo Fisher™ 42iTL NO_x trace analyser) were performed in outdoor air over 39 and 12 h, respectively. Results are reported in Fig. 3. The experiments took place at the Institute of Environmental Geosciences (IGE) in Saint-Martin-d'Hères, France. The IGE is located in the university campus, ∼1 km from the city centre of Grenoble and ∼300 m from a highway. Ambient air was pumped simultaneously from the same gas line by the instruments at flow rates of 1.0 and 0.5 L min⁻¹ for the IBBCEAS and the CLD instrument (Thermo Fisher™ 42i NO_x analyser), respectively. The measurements were conducted in September 2018. On the evening of Saturday, 29 September, the NO₂ peak occurred at a slightly later time than normally expected (from 20:00 to midnight; Fig. 3a). This may be due to the fact that on a Saturday night, urban traffic can be significant until late, but it also may be due to severe weather conditions prevailing at this time, with a storm and lightning known to be a major natural source of NO_x (Atkinson, 2000). For the second experiment shown in Fig. 3b, ambient air was pumped at flow rates of 1.0 and 0.8 L min⁻¹ for the IBBCEAS and the CLD trace instrument (Thermo Fisher™ 42iTL NO_x trace analyser), respectively. The measurements were conducted in July 2019. Both instruments showed the expected variability of an urban environment with an increase of NO₂ in the evening and morning due to photochemical processes and anthropogenic activities (i.e mainly urban traffic).

Figure 3(a) A 39 h long intercomparison of the IBBCEAS instrument and the commercial CLD instrument (Thermo Fisher™ 42i analyser) on the NO₂ detection in an outdoor urban area performed in September 2018 (dates in the format of two-digit year/month/day). The plot reports continuous (dashed lines) and hourly (dots) average data for both techniques. The grey area corresponds to the nighttime period. (b) A 12 h long intercomparison of the IBBCEAS instrument and the commercial CLD trace instrument (Thermo Fisher™ 42iTL trace analyser) on the NO₂ detection in an outdoor urban area performed in July 2019. The plot reports continuous (dashed lines) and 30 min (dots) average data for both techniques. The grey area corresponds to the nighttime period.

Download

The correlation plot, based on data of all instruments, in Fig. 4a, shows good linearity with a slope of 1.088±0.005 and a correlation coefficient of R²=0.989 with measurements averaged over 5 min. In order to perform linearity tests, the previous NO₂ FlexStream™ calibrator was used to produced various concentrations of NO₂ covering a large range of concentrations. Figure 4b shows the good linearity, from the ppt to ppb range, of the IBBCEAS instrument with a slope of 1.015±0.006 and a correlation factor of R²=0.9996, confirming the validity of the calibration approach. The discrepancies observed between the IBBCEAS and the CLD techniques might be explained by positive and negative interferences on the CLD technique. While the system presented here measures NO₂ directly, the CLD technique applies an indirect measurement of NO_x from the oxidation of NO through a catalyser, then in CLD; the NO₂ mixing ratio is obtained by the subtraction of the NO signal to the total NO_x signal. Villena et al. (2012) demonstrate that the interferences on an urban atmosphere for the CLD technique implied positive interferences when NO_y species photolysis occurred, leading to an overestimation of daytime NO₂ levels, while negative interferences were attributed to the VOCs photolysis followed by peroxy radical reactions with NO.

Figure 4(a) A linear correlation was obtained with a slope of 1.088±0.005 and a correlation coefficient of R²=0.989 between the IBBCEAS system and the Thermo Fisher instruments. (b) Results of the system's calibration using an NO₂ FlexStream™ calibrator. A linear correlation was obtained with a slope of 1.015±0.005 and a correlation coefficient of R²=0.9996.

Download

4.3 Instrument sensitivity and long-term stability

In remote areas such as East Antarctica, NO₂ ranges from a few tens to a few hundreds of ppt (50–300 ppt) (Frey et al., 2013, 2015). Due to the low signal-to-noise ratio of the spectrometer, a single acquired spectrum (with an integration time ranging between 200 and 350 ms) does not provide the required detection limit. However, the sensitivity can be improved by averaging the measurements for longer times, over which the instrument is stable. The stability of the IBBCEAS system is mainly affected by temperature fluctuations, mechanical instabilities and pressure drifts. In order to characterize the long-term stability of the instrument, two different studies were conducted on the IBBCEAS-NO₂ during the Antarctica field campaign at Dome C in 2019–2020 (the same results for the IBBCEAS-NO_x can be found in the Supplement). For both studies, the light transmitted through the cavity (I) was integrated at the CCD for 250 ms, providing a signal-to-noise ratio of 110 for a single spectrum. The reference spectrum (I₀) was taken by averaging 2000 individual spectra (∼8 min) while flushing the cavity with zero air. Subsequently, a 9 h long time series was recorded for each instrument maintaining the zero-air flow. The instrument was regulated at 12.0±0.2 ^∘C, with a cavity pressure of 630.0±0.7 mbar and a gas flow of 1.02±0.11 L min⁻¹. The minimum absorption coefficient (α_min), corresponding to the standard deviation of the residual of the spectrum, was deduced for different time averages. The results are shown in the log–log plot of Fig. 5, where the dots are the data and the dashed line indicates the trend in the case of a pure white-noise regime. From the graph one can see that the instrument follows the white-noise trend for about 22 min (5200 averages); afterwards, the baseline noise starts to deviate due to the rise of frequency-dependent noise. The chosen α_min value corresponds to $\mathrm{2.0}\times {\mathrm{10}}^{-\mathrm{10}}$ cm⁻¹ within ∼22 min (5200 spectra) of measurement during which a reference spectrum in the absence of absorbers and the absorption spectrum are acquired. The corresponding figure of merit (noise-equivalent absorption sensitivity – NEAS – or $\left.{\mathit{\alpha }}_{min\left(\mathrm{BW}\right)}={\mathit{\alpha }}_{min}\times \sqrt{\frac{t_{\mathrm{int}}}{M}}\right)$ is therefore $\mathrm{1.8}\times {\mathrm{10}}^{-\mathrm{10}}$ cm⁻¹ Hz${}^{-\mathrm{1}/\mathrm{2}}$ per spectral element (with t_int being the integration time of ∼11 min and M being the number of independent spectral elements; here 800 spectral elements are considered for the spectral fit).

Figure 5The minimum absorption coefficient α_min versus the number of spectral average for the IBBCEAS-NO₂ instrument. For these measurements the cell was continuously flushed with a flow of 1.02 L min⁻¹ of zero air, and the α_min was calculated from the standard deviation of the residual of the spectra at different time averages.

Download

For the same time series, an Allan–Werle (AW) statistical method on the measured concentrations was employed (Werle et al., 1993). In this case, spectra were averaged in a block of 10 and analysed by the fit routine. The results of the fit are reported (Fig. 6a). For an acquisition time of 2.5 s, corresponding to 10 averaged spectra, the AW standard deviation σ_AW-SD was 175, 4.8, 125 ppt and 725 ppb for NO₂, IO, CHOCHO and O₃, respectively. By increasing the integration time, the σ_AW-SD decreased following the white-noise trend, the coloured dashed line of Fig. 6b, with a characteristic $\sqrt{N}$ slope (where N is the number of averaged spectra). Because a reference spectrum in the absence of absorbers is required by this CEAS technique, depending on the shape of the AW plot, different strategies may be followed. In our case, the AW trends continue to decrease for all species for ∼22 min (5200 averages); this means that one can spend 11 min acquiring the reference spectrum and a further 11 min for the absorption spectrum, leading to limits of detection (LODs) of 11, 0.3, 10 ppt and 47 ppb (1σ) for NO₂, IO, CHOCHO and O₃, respectively. In our case, we chose to divide the measurement times by 2 (i.e. ∼11 min and 2600 averages for acquiring both the reference and the absorption spectra), offering equally interesting LODs: 16, 0.4, 12 ppt and 72 ppb for NO₂, IO, CHOCHO and O₃ (1σ), respectively, and allowing us to stay within the white-noise regime.

Figure 6(a) Mixing ratios of the target species NO₂, O₃, IO and CHOCHO measured during a 9 h Allan–Werle variance statistical experiment flowing zero air through the cavity on the IBBCEAS-NO₂ instrument. (b) The log–log Allan–Werle standard deviation plot, illustrating that the instrument performance follows the white-noise regime up to a certain extent, identified by the dashed lines. This represents the optimum integration time, after which instrumental instabilities start to dominate.

Download

The long-term stability and repeatability of the instrument were further studied by taking regular reference spectra within the optimum integration time of the instrument while continuously flushing the instrument with the zero-air mixture. In this case ∼5 and ∼3 min intervals were chosen, corresponding to 1024 and 580 averages (for 350 and 300 ms integration time, respectively) and a sensitivity on the NO₂ concentrations of 55 and 53 ppt (1σ), respectively. The results are reported in Fig. 7. Test₁ (1024 averages) was run for 9h, and Test₂ (580 averages) was run for 15 h. These tests highlight the reliability of the measurement protocol, with the long-term measurement well distributed within the 3σ level of the measurements precision (165 and 159 ppt, respectively). A boxplot is also reported representing the average values (green triangles) and medians, quartiles, and minimum and maximum values. The histogram analysis of those tests show a good distribution of the measurements within the 3σ level.

Figure 7(a, b) Time series of two long-term stability tests, with the black dashed lines representing the 3σ level; (c, d) boxplot of the stability test while continuously flushing zero air in the cavity, with means over 46 and 148 measurements (for Test₁ and Test₂, respectively) shown as green triangles and dots representing the outliers; (e, f) histogram analysis showing well distributed measurements within the 3σ level (black dashed lines).

Download

Table 1Performance comparisons with other recently developed IBBCEAS systems.

Download Print Version | Download XLSX

Table 1 shows a comparison between the instrument presented in this work and other recently developed IBBCEAS systems. The detection limits are given in ppt min⁻¹ (1σ) with the normalization time that accounts for the acquisition of the reference (without absorption) and sample spectra to allow for a better comparison. It should be noticed that all the other developments took advantage of an optical spectrometer with a cooled CCD device to reduce dark noise. A more compact and affordable spectrometer was preferred in this work. The cooling at the CCD would allow for a gain of up to a factor of 10 on the signal-to-noise ratio, which would directly apply to the achievable detection limits. Furthermore, a CCD with a higher sensitivity would allow for selecting higher-reflectivity mirrors and increase the optical pathlength. Noteworthy, the optimum integration time, corresponding to a minimum of the σ_AW-SD, is at 1300 s (∼22 min), allowing for the achievement of low detection limits even without a cooled CCD.

4.4 Indirect measurement of NO

Measuring NO and NO₂ simultaneously is important to study the NO_x budget in the atmosphere. In the selected blue visible region, there are no NO absorption features for direct optical measurements, and optical absorption detection of NO is typically done in the infrared region (Richard et al., 2018). However, its detection can be performed by indirectly measuring NO₂ after chemical conversion of NO to NO₂ in a controlled O₃ excess environment. This will lead to the measurement of NO_x, which, coupled with a simultaneous detection of NO₂ will provide the concentration of NO ($\left[\mathrm{NO}\right]=\left[{\mathrm{NO}}_x\right]-\left[{\mathrm{NO}}_{\mathrm{2}}\right]$) (Fuchs et al., 2009):

O₃ was produced by electrolysis of water using commercial ozone micro cells (INNOVATEC) allowing for the generation of O₃ without nitrogen oxide impurities and without the need of an oxygen gas bottle. The cells were mounted in a homemade plastic container offering a 200 cm³ water reservoir. With a miniaturized design (15 cm × 15 cm × 15 cm), ozone production can be controlled upon injection into the inlet line. The sample air flow to be analysed works as carrier gas for flushing the ozone-enriched surface water. This design also prevents the production of unwanted oxidizing agents such as peroxides, as well as sample dilution, causing a signal degradation and requiring precise flow measurements for quantitative analysis. The production of O₃ is controllable by the amount of electrolytic cells used and the supplied current, offering a dynamic range of 0–50 ppm of O₃ for a 1 L min⁻¹ total flow rate. A diagram and details of the system can be found in Sect. S3.2. For long-term use of the instrument, the overall water consumption should be considered. Losses due to evaporation were estimated to be between 7 and 30 cm³ d⁻¹ at 10 and 30 ^∘C, respectively, for a flow rate of 1 L min⁻¹, while losses due to electrolysis are negligible, with only 0.024 cm³ d⁻¹ of consumption. The other parameter to consider is the mixing time between the ozone generator and the measurement cell with respect to the O₃ excess. For instance, the calculated production rate of NO₂ from Reaction (R1) (i.e. reaction speed or conversion rate of NO) is $v=\mathrm{4.20}\times {\mathrm{10}}^{\mathrm{11}}$ molec cm³ s⁻¹ for 5 ppb of NO and 8 ppm of O₃. Under these conditions, a mixing time of 0.29 s is required for completing the conversion. With an air flow of 1 L min⁻¹, a 40 cm long 4 mm internal-diameter tube is therefore required between the ozone generator and the measurement cell. The performance of the ozone generation system was tested on the IBBCEAS instrument with a nitrogen oxide standard gas bottle containing ∼195 ppb of NO in air (Air Liquide). Kinetic simulations using Tenua software (Wachsstock, 2007) were made in order to establish the O₃ excess concentrations needed to achieve the complete conversion of NO to NO₂, which, along with its detection, was tested with the IBBCEAS instrument by varying the excess concentration of O₃ until complete conversion of NO was achieved at different flows (i.e. different reaction times before reaching the measurement cell). The experimental results were in good agreement with the simulations as reported in Fig. S9. In addition, the instrument was found to have a linear response regarding the detection of the produced O₃. The detection limit for the NO_x measurement was found to be similar to the one of NO₂, 10 ppt (1σ) in 22 min of integration time, while for NO, retrieved as the difference between the NO_x and the NO₂ concentrations, the detection limit estimated from the error propagation corresponds to 21 ppt.