Comparison of nitrous acid detection using open-path incoherent broadband cavity-enhanced absorption spectroscopy and extractive long-path absorption photometry

Comparison of nitrous acid detection using open-path incoherent broadband cavity-enhanced absorption spectroscopy and extractive long-path absorption photometry Sophie Dixneuf, Albert A. Ruth, (the late) Rolf Häseler, (the late) Theo Brauers, Franz Rohrer, and Hans-Peter Dorn 5 Department of Physics & Environmental Research Institute, University College Cork, Cork, Ireland. Institut für Energie und Klimaforschung, IEK-8: Troposphäre, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany. Current address: Bioaster Technology Research Institute Bioassays, Microsystems & Optics Engineering Unit, 40 Avenue Tony Garnier, 69007 Lyon, France 10 *Correspondence to: Albert A. Ruth (a.ruth@ucc.ie)


Introduction
Photolysis of nitrous acid (HONO + h (< 400 nm  OH + NO) leads to the formation of the most important daytime oxidant in the atmosphere, the hydroxyl radical (OH). Thus HONO indirectly affects the oxidative potential of the troposphere 25 and strongly influences degradation mechanisms of a vast variety of natural and urban pollutants. The mechanisms of HONO formation in the troposphere are still not fully understood (Calvert et al. 1994, Finlayson-Pitts et al. 2003, Ramazan et al. 2004, Liu et al. 2019. Many atmospheric studies revealed elevated HONO mixing ratios during daytime under specific conditions that cannot be fully explained (Staffelbach et al. 1997, Zhou et al. 2002a, Zhou et al. 2003, Vogel et al. 2003, Acker et al. 2006, Spataro and Ianniello, 2014. Although there appears to be general agreement 30 that heterogeneous NO2 chemistry is one of the most important sources of HONO (Harrison et al. 1994, Reisinger 2000, A long open-path broad-band cavity enhanced absorption instrument was set up at the SAPHIR 1 chamber at Forschungszentrum Jülich (Germany) in 2011, utilizing an incoherent short-arc Xe lamp as a light source (cf. Fiedler et al. 2003). Results of measurements to characterize its performance in the near UV through the in situ detection 65 of HONO and NO2, as well as methacrolein (MACR) are presented here. Similar to the recent work by Yi et al. 2021 our objectives in 2011 were (i) to assess the instrument's performance concerning HONO detection and validate it with the LOPAP system available at SAPHIR, and (ii) to compare its performance concerning NO2 detection with the local SAPHIR chemiluminescence spectrometry (CLS) NOX detector.
Section 2 outlines details on the experimental setup at SAPHIR in Jülich. Section 3 shows measurements of HONO,70 NO2 and MACR mixing ratio using the IBBCEAS instrument. The data are respectively compared with results from different instruments at SAPHIR: (a) a LOPAP system (HONO), (b) a standardized chemiluminescence detector (NO2), and (c) a protontransfer reaction mass spectrometer (PTRMS) (MACR). Correlation plots between individual measurements will be discussed for three measurement days in Summer and Autumn 2011. The experiments presented here supplement a campaign on the "Formal Intercomparison of Observations of Nitrous Acid" (FIONA) (Rodenas et al. 2013), where instruments for the detection 75 of HONO were compared at the EUPHORE simulation chambers in Valencia (Spain) in May 2009.

Experiment
IBBCEAS has been used for the detection of a variety of target species in different wavelength regions in the laboratory (Langridge et al. 2006, Washenfelder et al. 2008, Thalman and Volkamer 2010, Wu et al. 2009, Nakashima and Sadanaga 2017, Duan et al. 2018) and in outdoor environments (Bitter et al. 2005, Saiz-Lopez et al. 2006, Leigh et al. 80 2010, Wu et al. 2012. The high sensitivity and spatial resolution of open-path IBBCEAS make this approach particularly attractive for applications in atmospheric simulation chambers , Fuchs et al. 2010, Chen et al. 2011, Ashu-Ayem et al. 2012, Hoch et al. 2012, Rodenas et al. 2013. Thus an open-path IBBCEAS instrument was developed at the Jülich SAPHIR chamber, a facility designed for the simulation of tropospheric scenarios at realistically low mixing ratios of relevant trace species. The SAPHIR chamber is an ideal testbed for open-path IBBCEAS due to the available 85 comprehensive suite of standard detection methods that can be used to assess and validate the data taken with IBBCEAS. For the experiments presented here the IBBCEAS instrument was optimized for the near-UV detection of HONO (352-386 nm), whose identification was based on three prominent absorption bands peaking at 354, 368 and 384 nm .
The spectral region covered also enabled the simultaneous detection of NO2 and methacrolein.
The experimental design of the IBBCEAS instrument was similar to the one described in Refs. , Dorn et al. 90 2013 transmitter unit was mounted on a concrete platform at the north end of SAPHIR (facing south). The receiver unit was located at the south end of SAPHIR (facing north), in order to minimize exposure of the detector to potential daytime stray light. The dielectric mirrors (r = -21 m, R ~ 0.999, diameter = 40 mm, Layertec GmbH) inside the mechanically stable units formed an 95 optical cavity with a mirror separation of ~20.4 m. Each unit was temperature stabilized with an air-to-air thermoelectric assembly (Laird Technologies). Vertical metal plates with optical ports ( Figure S3) were rigidly mounted to the concrete support and formed the interface between the instrument's units and the chamber. Pipes connected to the two units extended into the SAPHIR chamber through the optical ports. Due to the use of the vertical plates (sealed by o-rings) the mechanical stress of the chamber's teflon wall was not imposed directly on the two optical units, which essentially decoupled the instrument 100 from the chamber wall and improved its long-term stability. Each unit had two ports: one pipe with 50 mm diameter and another one with 25 mm. In each unit the wider pipe was aligned along the optical axis of the cavity reducing the contribution of stray light to the measured signal in the receiver unit. The pipes were sealed with a high reflectivity mirror at one end and purged with zero air at a flow rate of 1.7 dm 3 min -1 during measurements to protect the mirror. The narrower pipe, which was sealed off during normal use of the instrument, was only used as a port for a green alignment laser (see Figure S1). The main 105 light source was a short-arc Xe lamp running in "hot-spot" mode. In this mode, a small plasma spot (~150 μm diameter) with a spectral radiance of 18 W cm −2 sr −1 nm −1 at 400 nm was formed close to the cathode surface, improving the imaging properties of the discharge in comparison to conventional diffuse arcs. Small random jumps of the arc position, however, required an active stabilization of the beam direction by means of a quadrant detector. When the spot "jumped" to a specific quadrant an actively controlled piezo-driven Al-mirror was used to minimise the resulting beam deviation from the optical axis of the 110 CEAS instrument. A telescope imaged the incoherent light into the optically stable cavity whose mirrors were aligned by two remote-controlled high resolution positioning motors per mirror. The light transmitted by the cavity was collected by a UVenhanced Al-mirror and focused onto the aperture of a circular-to-rectangular fiber bundle which guided the light onto the entrance slit (25 m) of a (Shamrock 303i) spectrograph (f = 303mm, F/4) supplied with a 1200 grooves/mm holographic grating. The light transmitted by the spectrograph was imaged onto a CCD detector cooled to -65°C. Light outside the high 115 reflectivity range of the cavity mirrors was optically filtered by means of a band-pass filter to avoid excessive scattering into the spectrometer and the potential saturation of the detector. The wavelength range from 352 to 386 nm was covered with a spectral resolution of ~0.24 nm. A sample of transmission spectra can be found in the supplementary material ( Figure S4).

Measurement Procedure
After overnight flushing of the chamber with dry synthetic air (also referred to as "zero air" in this publication) the cavity 120 transmission, I0(), was measured every morning in the dark chamber before experiments commenced. Typically 300 cavity transmission spectra with an individual acquisition time of 200 ms were accumulated during experiments, yielding a time resolution of 1 min per measured spectrum I(). The extinction coefficient () [cm -1 ] was calculated according to (the left part of) Eq. (1): (1) where Reff represents the wavelength dependent effective reflectivity of the cavity mirrors and d = 19.57 m the interaction path length per pass in the cavity (d equals the mirror separation minus the length corresponding to the mirror purge volume, see supplementary material, Figure S1). (1 -Reff) comprises all effective optical losses of the cavity when filled with zero air after 130 purging the chamber overnight; i.e. mirror reflectivity losses, Rayleigh (and potentially) Mie scattering losses, diffraction losses. The spatial average of the number density ni [molecule cm -3 ] of the absorbing species i (= HONO, NO2 and MACR) was retrieved by fitting the function on the right hand side in Eq. (1) to the measured extinction coefficient. i[cm 2 molecule -1 ] represent the absorption cross-sections of species i. The cross-sections for HONO , NO2 (Mérienne et al. 1995) and MACR (Meller et al. 1990) were converted to the spectral resolution of the spectrometer using home-made Gaussian 135 convolution software written in Fortran. The five fit parameters aj (j = 0...4) in Eq. (1) belong to a fourth-order polynomial, accounting for unspecified additional losses, such as background featureless absorption, or Rayleigh and Mie scattering that may become relevant over time. The fitting procedure was based on least squares minimization using a singular value decomposition (SVD) procedure (Press et al. 1986) in order to eliminate biases of the fit due to parameter correlations. During a first test run of the SVD approach the wavelength of all cross-section reference spectra were individually 140 shifted to further minimize the least square sum, and the optimized wavelengths were subsequently used in SVD analyses. For all reference spectra the typical shift was ≈0.10  0.05 nm and thus within the spectral resolution of the measurement. The absolute wavelength (e.g. used in Fig. 4) was calibrated with a low-pressure neon pen ray lamp. Reflectivity calibration issues will be discussed in the section 4.

The LOPAP Instrument 145
The LOPAP instrument used in this study has been described in detail by Häseler et al. 2009 andLi et al. 2014 in the context of several campaigns on airborne detection of HONO aboard a Zeppelin airship. Air is extracted through two sampling coils in series where HONO is stripped into the liquid phase. In both coils the air is exposed to equal flows of a solution (0.06 M sulphanilamide in 1 M HCl solution) in which HONO almost instantaneously reacts to forms a diazonium salt. While the first coil removes HONO nearly quantitatively from the gas phase, only a fraction of other chemically interfering species are 150 scrubbed. The second coil, however, samples that fraction of interfering species but only the remaining small amount of HONO. Using the difference of the signals derived from the two coils enables the influence of interfering species to be accounted for. The air is then separated from the liquid and the solutions are separately transferred into two mixing volumes, where a 0.8 mM solution of N-(1-naphthyl)ethylenediamine-dihydrochloride is added to generate an azodye. The azodye's concentration is then determined by its optical absorption to determine HONO mixing ratios. The absorption cells for both 155 https://doi.org/10.5194/amt-2021-291 Preprint. Discussion started: 22 November 2021 c Author(s) 2021. CC BY 4.0 License. channels consist of long length teflon tubing, acting as a liquid core waveguide (LCW). Visible light is sent through the tubing and detected by two small spectrometers.
The entire instrument was housed in a compact 19" rack (56(w) × 60(d) × 100(h) cm 3 ) consisting of two "chemistry" units and one "electronics/detection unit". In order to avoid sampling artefacts in inlet lines the sampling unit was straightforwardly mounted inside SAHPIR about 0.3 m from the chamber wall at the north end and 1 m above the floor. The instrument's 160 sampling frequency, time resolution, 3 detection limit, 1 precision, and accuracy of the instrument was 0.33 Hz, 4-5 min, 10 pptv, 3 pptv, and 12 %, respectively. The measurement range of the instrument can in principle be varied by the length of the absorption tubes and by the use of different absorption wavelengths for the evaluation. In this study we used an optical path length of 2.9 m.

Chemiluminescence Spectrometry NOX Detector 165
The NOX detector was located in a container underneath the chamber from where gas mixtures were sampled at a flow rate of 1 dm 3 min -1 through a teflon tube of ca. 6 m length (internal diameter 4 mm) corresponding to an approximate residence time of 1 s. NO2 was converted to NO by a pulsed LED photolytic converter at 3958 nm (Droplet Measurement Technologies, BLC) in a volume of 17 ml with a conversion efficiency of approximately 50%. The LED in the converter was consecutively switched on and off to alternately determine NO and NOX concentrations. NO was detected by a customized CLS detector 170 (Eco Physics TR 780 Brüning 1992, Fuchs et al. 2010). NO2 mixing ratios were calculated using an interpolated value between two subsequent NOX measurements at a time when NO mixing ratio were measured. The instrument was calibrated using NO standard gas mixtures (2 ppmv NO in N2, BOC-Linde) and gas phase titration for NO2.
The fact that besides NO2 also HONO is photolysed at 395 nm to yield NO was accounted for by determining the corresponding NO yield from HONO numerically from the spectrum of the LEDs. The HONO photolysis contribution to NO is less than 5% 175 compared to that of NO2. The 1 accuracy of the chemiluminescence detector for NO2 was determined to be ±7%, based on the uncertainty of ±5% of the NO standard used for the calibration and a ±5% uncertainty for the NO2 conversion efficiency.
The known interference of 5% towards HONO is not corrected in the final dataset and not included in this accuracy estimate.

Proton Transfer Reaction -Mass Spectrometry (PTRMS)
PTRMS was utilized to monitor methacrolein in the presence of HONO and NO2. Generally the PTRMS technique relies on 180 soft chemical ionization to detect gaseous trace components. The target species are converted to product ions through the transfer of a proton from the reagent ion, H3O + . The trace gases (X) are identified through the mass of the product ions usually being the protonated molecular mass (XH + ): H3O + + X  XH + + H2O. The PTRMS instruments applied here was a quadrupole mass spectrometer system (PTR-Quad-MSThe system features a switchable reagent ion source with H3O + , NO + and O2 + as precursor ions for the measurement and identification of a number of trace gases. Details on the PTRMS instrument were 185 published earlier by Wisthaler et al. 2008.

Results
Measurements of time-dependent mixing ratios of HONO, NO2 and MACR using open-path IBBCEAS were taken during Summer and Autumn 2011, and compared with those utilizing long-path absorption photometry (LOPAP), chemiluminescence spectrometry (CLS) and proton transfer reaction mass spectrometry (PTRMS), respectively. The performance intercomparison 190 study is exemplified on basis of measurements on the 11 th of July and on the 5 th & 6 th of October 2011, when different photochemical scenarios were simulated. The measurements in July were carried out as part of a Jülich internal photochemistry campaign (6 th June to 15 th July 2011), whose goal was to study the oxidation of isoprene (H2C=C(CH3)-CH=CH2), methacrolein CH2=C(CH3)-CHO or methyl vinyl ketone (CH3-C(O)-CH=CH2) by hydroxyl (OH) radicals at low NOX mixing ratios . In contrast, the measurements in October were specifically designed for a 195 comparison between LOPAP and IBBCEAS under well controlled low concentration conditions with no obvious potential chemical interferences disturbing the LOPAP instrument. Generally, after cleaning and humidifying the SAPHIR chamber HONO formation by unknown photo-induced reactions on the Teflon chamber walls and degradation was studied in experiments always including light-induced and dark reactions of HONO formation or destructionthe experimental protocols concerning changes in chamber conditions are given in the figure captions. The results obtained on the three days will be 200 outlined in chronological order. The performance of the open-path IBBCEAS instrument will subsequently be discussed in the context of the different measurement condition and atmospheric scenarios together with that of the Jülich LOPAP instrument.  Figure 1 indicates the times when the cavity transmission through the clean chamber, I0(λ), was measured for ca. 10 min. 11 July is the only day for which the build-up of HONO was monitored during daylight conditions. The HONO mixing ratio increased after humidification of the bright chamber (for 48 min) and decreased subsequently when O3 (~40 ppbv) was introduced. At 9:15 hrs there is a marked but unexplained change in the data of HONO mixing ratios as 210 measured by IBBCEAS in comparison to LOPAP. The increase in the noise of the IBBCEAS data occurred ca. 15 minutes before the addition of CO to the chamber at 9:30 hrs, which is neither expected to influence the HONO chemistry nor the data retrieval, even at high CO (~750 pbbv) concentrations (see Figure 1). Presently there is no obvious explanation for this behavior in the IBBCEAS measurement. Likewise the return to dark conditions at 15:33 hrs led initially to an unexpected increase of the HONO mixing ratio as recorded by the LOPAP instrument, but it was also observed by the IBBCEAS measurement. This 215 observation was also made in other campaigns and will be briefly discussed in section 4.2. NO2 mixing ratios, as measured by CLS, increased gradually during humidification of the bright chamber and increased sharply when O3 was introduced, followed by a more gradual increase until MACR was added to the mixture at 11:51 hrs. The jump of the NO2 mixing ratio at 07:30 shows the effect when O3 was added actively to the chamber. Almost all NO was oxidised to NO2 at that time. Both CLD and CEAS show this effect. The observed temporary decrease of the NO2 concentration after the addition of MACR can be 220 explained by a reaction sequence following the formation of OH from HONO: The primary oxidation step of the reaction of OH with MACR forms the peroxy methacryloyl radical (CH2=C(CH3)-C(O)OO), part of which further reacts with NO2 to form MPAN (peroxy methacryloyl nitrate, CH2=C(CH3)-C(O)OONO2). MPAN is thermally unstable with a small thermal decomposition rate of ~ 4.6×10 -4 s -1 at 25 o C (Roberts and Bertman 1992). Thus a thermal equilibrium is established from which NO2 is reformed after the initial MACR has been consumed . 225 Concentrations of NO2 and MACR both appeared to stagnate when the roof was closed due to the absence of light-driven photo-chemistry. Finally the purging of the chamber with synthetic air removed all trace gases from the chamber.
The experiments in October were designed to enable accurate mirror reflectivity calibration and intercomparison of LOPAP and IBBCEAS under unperturbed condition at sub-ppbv mixing ratios of HONO. The measurements demonstrate that the long-cavity IBBCEAS instrument at SAPHIR is capable of detecting HONO pptv levels without difficulty (Figures 2 and 3). 230

Measurements on 5 October 2011
After overnight flushing in the morning of 5 October, NO2 was added to the chamber in steps of 250 pptv (at 6:30 hrs and 6:50 hrs), 500 pptv (at 7:10 and 7:30), and 1 ppbv (at 7:50 hrs and 8:10 hrs). After subsequent humidification for 44 min (starting at 8:41 hrs) and the exposure to daylight at 9:27 hrs, HONO was formed at levels of up to 400 pptv. Finally, after closing the chamber roof (14:23 hrs), HONO was removed by flushing with zero air (starting at 15:31 hrs). The HONO data 235 from the IBBCEAS and LOPAP setups, and the NO2 data from the IBBCEAS and CLS instruments showed outstanding agreement on that day (Fig. 2).

Figure 2 3.3 Measurements on 6 October 2011
It is known that the photo-enhanced formation of HONO in the SAPHIR chamber can be described by an empirical function 240 depending on relative humidity, solar irradiation and temperature (Rohrer et al. 2005). Minimal HONO production was thus achieved by only humidifying the chamber for 42 min after overnight flushing with zero air. At 12:18 hrs the chamber was exposed to daylight leading to the gradual formation based on vestiges of NOX on the chamber wall. The variation of temperature was limited to the natural variability. After a gradual increase the HONO mixing ratios leveled off at ca. 250 pptv before the chamber was eventually closed at 16:24 hrs. The correlation between the data obtained with IBBCEAS and LOPAP 245 is rather satisfactory at these low levels. For NO2 there appears to be a slight offset between data from the IBBCEAS and CLS instruments with good overall agreement. The main experimental uncertainties determining the quality of the IBBCEAS data reported here are systematic; they are: (a) the stability of the light source, (b) the in situ calibration of the mirror reflectivity, (c) the data analysis and concentration retrieval approach, and (d) the unspecified mechanical instabilities of the setup such as potential thermal drifts or deficiencies of the opto-mechanical components (Ruth et al. 2014). The latter become more critical with increasing cavity length since small changes in the optical alignment have a more severe influence on the instrument's performance; see Varma et al. 2009. 255

Lamp Stability
Short-term intensity fluctuations are due to random spatial variations of the hot spot plasma arc (small arc jumps), whose effects on the optical alignment were minimised by an active quadrant detector control unit ). As long as intensity fluctuations of the lamp do not show any spectral dependence in the wavelength range of interest, the resulting baseline changes can in principle be accounted for in the fit of eq (1) to the measured data. If the fluctuations are however 260 accompanied by random spectral variation, the retrieval by SVD becomes increasingly difficult. Spatial jumps that were at the compensation limit of the quadrant detector were also noted during experiments leading to uncertainties in the baselines. One of these events may have occurred e.g. on 11 July at ~9:30 hrs. Difficulties in the mixing ratio retrieval due to lamp instabilities occurred sporadically during the measurements in the Summer, but became more frequent and for longer time periods after continued use of the lamp in the Autumn. For instance the measurement on 6 Oct (Figure 3) was affected by lamp stability 265 issues, which is a sign of electrode aging and the main cause for the hot spot plasma arc to wander into a domain where the stabilization system is unable to fully compensate for the spatial displacement.
An unexplained change in performance started approximately 20 min before the addition of 750 ppbv of CO on 11 July at ca. 9:10 hrs (Figure 1). At that time the quality of the IBBCEAS data started to worsen, while the LOPAP performance remained largely unaffected (see Figure 1). A small increase in HONO mixing ratios was subsequently observed in comparison to the 270 LOPAP and the noise of the IBBCEAS data was increased by roughly a factor of ~2. There are no obvious reasons for this observed behavior from a (photo)chemical point of view, since at the time the gas mixture was not altered. In open path cavity setups similar behavior can in principle be caused through increased scattering due to particle formation ), but measurements of the particle number concentration showed no significant change on this occasion. Therefore one conceivable explanation may be that the hot spot arc in the Xe lamp may have moved to a steady location on the electrode 275 where the control unit was at its performance limit to keep the arc steady. As the quadrant detector correction signal was not recorded, this is merely a tentative explanation of the sudden change of performance. Although there was an addition of CO a short time later it is not plausible that potential impurities in the CO (purity >99.9%) gas (or chemical reactions of same) might be the cause for this observation. LED-based IBBCEAS is generally less prone to sudden changes (Gherman et al. 2008) due to the higher stability of the light source, although slow drifts are still possible (Fouqueau et al. 2020).

Mirror Reflectivity and Calibration Aspects
Retrieval of accurate mixing ratios by IBBCEAS requires the effective mirror reflectivity Reff to be known accurately as a function of wavelength. Calibration measurements were performed a few times over the course of the measurements presented here by introducing a known amount of NO2 into the dry SAPHIR chamber shortly after the cavity transmission of the clean and dry cavity, I0(), had been recorded. The effective reflectivity was retrieved with eq (1) using the known cross-section of 285 NO2 and the mixing ratio from the CLS NOX monitor (e.g. using the measurements on 5 th Oct 2011 where Reff = 0.9978 at 352 nm and Reff = 0.9986 at 386 nm, see supplementary Figure S5). The lower limit of the absolute uncertainty (~11%) is based on the accuracy of the NO2 cross-section (8%) and that of the NOX measurement (7%).
The initial calibration measurement of Reff was also used to determine the loss, LLLO(), of an anti-reflection coated optic (referred to as "low loss optic", LLO), which was in turn used on a daily basis (in the morning) to determine the reflectivity in 290 the clean chamber instead of using NO2 as calibration gas (see also supplementary Figures S4 and S6): In the transmitter unit the LLO (RLLO (375 nm) < 0.001, diameter 40 mm, parallelism 30'') was moved into and out of the cavity 295 parallel to the mirror by means of an accurate translational stage with high reproducibility. Care was taken in designing the air-tight LLO compartment ( Figure S1) where the optic was "parked" when not needed. The LLO was furthermore flushed with clean dry air while in the cavity to avoid (minimize) potential changes of its optical loss. When the LLO is used over the course of a simulation experiment, i.e. in a chamber with arbitrary gas mixture, the LLO measurements yield effective reflectivities that comprise extinction losses in the chamber at this particular time. Although only the Reff measured each 300 morning in the clean dry and dark chamber was used to retrieve the concentrations of the target species (eq (1)) for the entire day, repeated introductions of the LLO into the cavity over the course of the day was used for checking whether misalignments or potential drifts of the cavity had occurred.

Data Evaluation 305
In order to judge the quality of the data evaluation an example of an IBBCEAS extinction spectrum and the corresponding fit of eq (1) to the measured data are shown in Figure 4 (uppermost panel). The data for this example were chosen because on 11 July 2011 HONO, NO2 and MACR were simultaneously present in the chamber. The different panels in Figure 4

Detection Limits
For individually measured spectra a minimum extinction coefficient of 9.310 -10 cm -1 was determined for an acquisition time of 1 min (assuming a 1:1 signal-to-noise ratio), which corresponds to min = 7.210 -9 cm -1 Hz -1/2 (1 For a series of measurements (see supplementary Figure S7) this translates into a measured 3 limit of detection (LOD) of the IBBCEAS 315 instrument for a 1 min acquisition time of ~39 pptv for HONO, ~114 pptv for NO2, and ca. 510 pptv for MACR in the 352-386 nm wavelength range.
In comparison to the IBBCEAS instrument, the LOPAP instrument at SAPHIR featured a 3 LOD of 10 pptv of HONO in 1 min. The LOD for NO2 by the standardised chemiluminescence technique at SAPHIR also compares favourably with a 3σ LOD of 13.4 pptv of NO2 for a 1 min acquisition time with an overall accuracy of 7 % (Rohrer and Brüning 1992, Fuchs et al. 320 2010) (the CLS detection limit for NO is 6.7 pptv). Mixing ratios of volatile organic compounds such as methacrolein can be only compared to data acquired with PTRMS, whose LOD for MACR is 8.5 pptv in 1 min (accuracy 8 %).
An overview of (2) detection limits for HONO of previously published IBBCEAS instruments was recently given by Jordan and Osthoff 2020. The 2 detection limit of 26 pptv for HONO in 1 min presented here compares favourably to the IBBCEAS works by Gherman et al. 2008, Hoch et al. 2012, Wu et al. 2012, Donaldson et al. 2014, Scharko et al. 2014, Min et al. 2016, 325 Nakashima and Sadanaga 2017, Duan et al. 2018, and Jordan and Osthoff 2020 -see Table 1 in the AMT article by Jordan and Osthoff 2020. It should be noted, however, that some of the quoted detection limits were established under field conditions, which bears challenges that are different from experiments with a large scale simulation chamber. The rather competitive detection limits of the current setup are a result of the substantial cavity length, despite the fact that the effective cavity mirror reflectivity in this work was lower than in all other instruments reported. 330

Figure 4 4.2 LOPAP Instrument
In a previous DOAS-LOPAP intercomparison ) the addition of ozone triggered issues concerning the use of proper reference data in the DOAS evaluation procedure at ppbv levels of HONO. This sort of difficulty was not observed with IBBCEAS at sub-ppbv mixing ratios of HONO. 335 However, when switching from illuminated to dark conditions the LOPAP instrument regularly showed an increase of HONO concentrations, while the IBBCEAS detector does not seem to follow this trend as strongly and reproducibly; see Figure 1 (11 July 2011) or Figure S8, for example. The systematically increasing HONO mixing ratios measured by the LOPAP instrument upon closing the roof and generating dark conditions was also observed on a few other days during the summer campaign, e.g. on 10 and 15 June (see supplementary Figure S8). The reason for this behavior is not fully understood yet. An effect of the 340 presence of organic reaction products from the oxidation of mathacrolein that cause an interference with the LOPAP technique upon stopping photolytic processes can be ruled out, because the "effect" was also observed in experiments with a clean humidified chamber solely filled with high purity synthetic (zero) air. A possible explanation may be based on the changing https://doi.org/10.5194/amt-2021-291 Preprint. Discussion started: 22 November 2021 c Author(s) 2021. CC BY 4.0 License.
homogeneous and heterogeneous production rates of HONO versus its destruction rates under bright and dark conditions. For daytime conditions the steady state HONO mixing ratio is mainly determined by photolysis of a yet unknown nitrogen 345 containing precursor adsorbed at the Teflon walls or dissolved in the surface water layer on the Teflon film versus the photolytic destruction of HONO into OH and NO. In the dark (closed) humidified chamber HONO can still be produced, but predominantly through heterogeneous reactions with formation rates being at least a factor of ~10 smaller than under illuminated conditions (i.e. with the roof open). Even though closing the roof stops efficient HONO production it also eliminates the photolytical loss entirely, while HONO heterogeneous production within the aqueous phase on the chamber 350 walls is still active (albeit its slowly decreasing efficiency). This can lead to changing the equlibrum such that the subsequent outgassing of HONO from the aqueous phase on the Teflon film can temporarily dominate the HONO production into the gas phase until the initially HONO-saturated aqueous phase also gets depleted leading to the eventual decrease in HONO mixing ratios (Karl, 2004). The inlet of the LOPAP instrument is much closer to the chamber wall (ca. 30 cm) in comparison to the region probed by IBBCEAS, which averages across the whole length of the chamber near its center. Thus the effect is more 355 likely to be apparent in the LOPAP data than in the IBBCEAS data. This tentative explanation of the observed trend upon closing the roof of the chamber warrants further investigation.

CLS and PTRMS Instruments
The CLS and PTRMS data were used here as reference guideline for the measurements and not for scrutinizing the performance of the respective instruments. Nevertheless, some discrepancies between IBBCEAS, CLS and PTRMS were noted and are 360 outlined here.

CLS:
Mixing ratios of NO2 measured by CLS and IBBCEAS frequently appeared to differ somewhat during the humidification of the dark chamber. The CLS instrument appears to systematically record higher NO2 mixing ratios than IBBCEAS upon humidification of an initially clean chamber without the obvious presence of NO2. Examples of the observed behavior are 365 shown in Figure 5 (see also Figure S8).

Figure 5
There are two possible scenarios that may explain these observations: (a) It could be an artefact in the chemiluminescence device, which may be due to a surface effect of the blue light converter (BLC). Switching from a dry to a humidified chamber leads to a severe change in the surface conditions of the BLC, which releases either NO or a substance that mimics an NO 370 signal for some time until a new equilibrium has been reached. This type of behavior has occurred before in measurement campaigns at SAPHIR, where small temporary discrepancies between model calculations and measurement were observed.
The nature of the potentially released substance is unclear. HONO would be more likely to stick to the walls rather than to be released unless heterogenous reactions are at play leading to the formation of NO. The effect is transient and vanishes after a short time. The fact that rapidly changing water vapor concentrations can affect the instrumental background of CLS detectors 375 and lead to a non-trivial memory effect that cannot be easily corrected retrospectively has also been reported recently in the https://doi.org/10.5194/amt-2021-291 Preprint. Discussion started: 22 November 2021 c Author(s) 2021. CC BY 4.0 License.
literature (Nussbaumer et al. 2021). (b) A small contamination of the milli-Q water and differences in the sample quality used for humidification cannot be fully ruled out. Those contaminations have also been observed before. The increase in the NO2 signal upon humidification is supported by a significant increase in the NO signal (~120 pptv) observed by CLS. However, on 10 June 2011 this increase seems to be over after ca. 15 min while the humidification is still ongoing. The reason why the 380 IBBCEAS instrument does not detect the increase in NO2 if it is due to contamination is also unclear.

PTRMS:
After introduction of MACR to the chamber on 11 July, initially higher mixing ratios were recorded by PTRMS rather than by IBBCEAS (between 11:50 hrs and ca. 15:30 hrs). The mixing ratios measured with the two instrument gradually approach very similar values over the said time period until they agree rather well when the roof of the chamber was closed. 385 The reason for this behavior is not clear since no obvious interference in this period was obvious.

Correlation of Data Obtained by Different Instrumental Methods
For the combined data sets of the instrument intercomparisons an overall good agreement for both trend and absolute mixing ratios was observed between IBBCEAS and the established instruments at SAPHIR as illustrated in the correlation plots in 390 Figures 6 and Table 1 (see also Figs. 1-3). The best agreement is observed on 5 Oct 2011 for both HONO and NO2, which may be attributed to the fact that the least amount of interfering species were present in the chamber at that time. R coefficients range from 0.930 to 0.994 for HONO and between 0.937 and 0.992 for NO2. An R coefficient of 0.962 is found for the single measurement of MACR.

Conclusion 395
In 2011 an incoherent broadband open-path cavity-enhanced absorption spectroscopy (IBBCEAS) instrument was established at the SAPHIR chamber in Jülich and optimized for the detection of HONO and NO2 in the near UV region (352-386 nm) using a bright hot-spot Xe-arc lamp and a UV-enhanced CCD detector. Based on an effective reflectivity of 0.9978 < Reff < 0.9986 and a 20 m open-path cavity a 2 detection limit of 26 pptv for HONO and 76 pptv for NO2 was achieved for an integration time of 1 min. Methacrolein was also detected at mixing ratios below 5 ppbv. These are highly competitive detection 400 limits in comparison to those reported in the recent literature. The IBBCEAS instrument's performance for HONO and NO2 detection was compared to that of long-path absorption photometry (LOPAP) and chemiluminescent NOX detection, with R coefficients range from 0.930 to 0.994 for HONO, and between 0.937 and 0.992 for NO2, respectively. At low concentrations however, and especially upon humidification of the chamber, a small temporary water interference on NO2 mixing ratios were observed in CLS measurements which was not observed in the IBBCEAS data. Tillmann, for providing the PTRMS data for this publication. The extraordinary craftsmanship of the mechanical workshop at Jülich in the context of this project is also gratefully recognized.  N: number of data points, a and b: regression coefficients, σ (a) and and σ (b)

695
The vertical black arrow indicates the time when I0 was measured (generally for 10 min).