A five-channel cavity ring-down spectrometer for the detection of NO 2 , NO 3 , N 2 O 5 , total peroxy nitrates and total alkyl nitrates

We report the characteristics and performance of a newly developed five-channel cavity ring-down spectrometer to detect NO3, N2O5, NO2, total peroxy nitrates (6PNs) and total alkyl nitrates (6ANs). NO3 and NO2 are detected directly at 662 and 405 nm, respectively. N2O5 is measured as NO3 after thermal decomposition at 383 K. PNs and ANs are detected as NO2 after thermal decomposition at 448 and 648 K. We describe details of the instrument construction and operation as well as the results of extensive laboratory experiments that quantify the chemical and optical interferences that lead to biases in the measured mixing ratios, in particular involving the reactions of organic radical fragments following thermal dissociation of PNs and ANs. Finally, we present data obtained during the first field deployment of the instrument in July 2015.


Introduction
Nitrogen oxides play a central role in tropospheric chemistry though their influence on ozone production and radical cycling, and thus on the oxidative capacity of the atmosphere.Nitric oxide (NO) is directly emitted into the boundary layer by e.g.fossil fuel combustion and biomass burning, and is oxidised via various mechanisms (e.g.reaction with O 3 and peroxy radicals, RO 2 ) to form NO 2 , which is the dominant precursor of photochemically formed, tropospheric ozone.Together, NO and NO 2 are referred to as NO x , while the sum of all reactive, oxidised nitrogen species (NO x plus inorganic and organic nitrates and nitrites and halogenated nitrogen oxides) is referred to as NO y .In the tropospheric boundary layer, NO 2 mixing ratios vary from < 10 pptv in remote re-gions to > 100 ppbv in highly polluted environments.Tropospheric NO 2 has been measured using several techniques including differential optical absorption spectroscopy (DOAS), chemiluminescence (CLD), laser-induced fluorescence (LIF) and cavity ring-down spectroscopy (CRDS).For a summary and intercomparison of these methods, the reader is referred to Fuchs et al. (2010).NO 2 is the major precursor of the inorganic nitrates NO 3 and N 2 O 5 .During the daytime, NO 3 is rapidly photolysed (Johnston et al., 1996) and also reacts with NO to reform NO 2 so that NO 3 mixing ratios are very low.At night, NO 3 levels may reach hundreds of pptv in polluted air, and their reactions represent a significant route for oxidation of volatile organic compounds (VOCs) and, via formation of N 2 O 5 , are a source of particulate nitrate and thus sink of NO x (Wayne et al., 1991;Allan et al., 1999).Tropospheric NO 3 has been detected using long-path DOAS (LP-DOAS; Platt et al., 1980) over a path length of typically > 1 km, whilst point detection methods of NO 3 include cavity-enhanced DOAS-, LIF-and CRDS-based instruments.N 2 O 5 has been detected as NO 3 after thermal decomposition or directly measured by chemical ionisation mass spectrometry.Descriptions of methods for detecting NO 3 and N 2 O 5 and an intercomparison experiment have recently been published (Fuchs et al., 2012;Dorn et al., 2013).
Reactions between atmospheric oxidants such as OH or NO 3 and anthropogenic or biogenic volatile organic compounds (VOCs) lead to the formation of organic peroxy radicals that react with NO or NO 2 to form organic nitrates (Roberts, 1990).Reaction of RO 2 with NO 2 results in the formation of peroxy nitrates (RO 2 NO 2 , henceforth abbreviated as PNs).If the initial peroxy radical contains a terminal acyl group (R(O)O 2 ), peroxyacetic nitric anhydrides (also called peroxyacyl nitrates or PANs, R(O)O 2 NO 2 ) are formed, which have lifetimes on the order of hours in temperate boundary layer conditions but weeks in cold regions of the atmosphere.Following generation and transport away from pollution sources, PANs can thermally decompose to NO 2 and an organic radical, thus providing NO x and enhancing the production of O 3 in remote locations.Non-substituted peroxy radicals react to form PNs that are less thermally stable, which undergo rapid (timescale of seconds at 298 K) thermal decomposition in the mid-latitude boundary layer, and thus only reach higher concentrations in colder regions such as the upper troposphere and regions close to the poles (Slusher et al., 2002;Murphy et al., 2004).The reaction of RO 2 with NO mainly forms RO + NO 2 , but can also result in the formation of alkyl nitrates (RONO 2 , henceforth abbreviated as ANs) in a minor branch (1-30 %) of this reaction, with higher yields for long-chain peroxy radicals (Arey et al., 2001;Perring et al., 2013).ANs are also formed at night from the reaction between NO 3 and VOCs with a higher yield (20-80 %) than the OH-initiated process (Perring et al., 2013).The oxidation of alkenes by NO 3 results in the formation of nitroxy alkyl radicals that can react with O 2 to form nitroxy alkyl peroxy radicals that can undergo further chemistry.ANs are longer lived than PANs against thermal decomposition, and can thus also act as NO x reservoirs.Reactions with OH or photolysis or transfer to the aerosol phase are the major sinks of ANs, the relative importance depending on carbon chain size and degree of substitution.PNs and ANs represent a very wide group of species present in the atmosphere, with different levels of structural complexity due to the different possible combinations of functional groups.Individual nitrates can be measured using chromatographybased instruments followed by electron capture detection, conversion to NO or NO 2 (Roberts et al., 1989;Roberts, 1990;Flocke et al., 1991;Blanchard et al., 1993) or chemical ionisation mass spectrometry (Beaver et al., 2012), whereas the detection of the total (i.e.non-speciated) ambient PNs and ANs has been demonstrated using thermal decomposition of the nitrates followed by NO 2 detection (Day et al., 2002(Day et al., , 2003;;Paul et al., 2009;Paul and Osthoff, 2010;Wooldridge et al., 2010;Thieser et al., 2016).
A single instrument that measures NO 2 , NO 3 , N 2 O 5 , total peroxy nitrates ( PNs) and total alkyl nitrates ( ANs) can clearly provide great insight into the reactive nitrogen budget of the troposphere, enabling studies of daytime and night-time processes that transfer NO x from the gas to the particle phase or to organic reservoir species.The interconversions and relation to NO x losses and photochemical O 3 generation of these traces gases are described in Fig. 1, and those highlighted in red circles are detected with this instrument.The five-channel instrument we describe is a further development and extension of the two-channel, onecolour (405 nm) prototype recently described by Thieser et al. (2016) and of the two-channel set-up for measuring NO 3 and N 2 O 5 previously developed and operated by this group  (Schuster et al., 2009).Whilst some basic concepts are similar, several significant design modifications (optical, mechanical, electrical/data-acquisition and chemical) have been made that improve the precision, the limit of detection and the accuracy of the measurements.These improvements are described in the relevant sections below.
In Sect. 2 we describe the overall design of the instrument and the features specific to the measurement of the individual trace gases.In Sect. 3 calibration procedures and laboratory characterisation experiments are described, along with an assessment of the overall uncertainty.Finally, in Sect. 4 we show data from a first field deployment at a semi-rural mountain site in Germany.

Instrument design
The instrument we describe uses cavity ring-down spectroscopy (CRDS), partly coupled with thermal dissociation (TD-CRDS).CRDS is a detection technique that was developed in the early 1990s (O' Keefe and Deacon, 1988;Lehmann and Romanini, 1996), and has been recently reviewed (Berden et al., 2000;Brown, 2003).The underlying principle of CRDS is the measurement of optical extinction in a (usually) closed cavity.In contrast to differential absorption spectroscopy in which optical extinction is linked to a measured light intensity difference, in CRDS set-ups, the optical extinction is calculated from the rate of decay of light intensity (ring-down constant) measured at the output of the cavity after a light pulse (pulsed CRDS) or after a continuous light source is turned off (cw-CRDS).The light intensity decay, measured behind a cavity mirror after switching the light source off, is exponential, and the concentration of an absorbing/scattering gas (i) can be calculated using Eq. ( 1): where τ 0 and τ (s) correspond to decay constants in the absence and presence of absorbing/scattering trace gases, respectively (see Sect. 2.1 and 2.2). σ (i,λ) is the absorption cross section/scattering coefficient of species i at wavelength λ.The determination of the decay constants is done using a linear regression of sums fitting method (LRS), which has a number of advantages compared to least-squares (LS) fitting (Everest and Atkinson, 2008), including a faster algorithm and the removal of the need to measure an accurate zero (or baseline), which enables use of shorter on-off modulation cycles and higher data acquisition frequencies.Comprehensive comparison between the LS and LRS methods for both synthetic and real ring-down decays showed excellent agreement.
The central part of the instrument described here consists of five cavities of similar design and materials (see Fig. 2).Two cavities are used to detect NO 3 and N 2 O 5 (at 662 nm), and three are used to detect NO 2 , PNs and ANs (at 405 nm).We refer to the five different inlet/cavity combinations (downstream of the filters, see Fig. 2) as the NO 2 , NO 3 , N 2 O 5 , PNs and ANs channels.The light sources for the five cavities are two laser diodes housed in Thorlabs TCLDM9 modules located in aluminium hous-ing with optical components for beam splitting, collimation and optical isolation (see Sect. 2.1 and 2.2).Coupling between the laser diodes and the cavities is achieved using 50 µm core optical fibres with collimators adjusted to weakly focus the beam (diameter of ≈ 3 mm) at a point ≈ 1 m behind the cavity output mirrors.The cavities are constructed from 70 cm long, seamless stainless-steel pipes (1 mm thick, with an inner diameter of 8 mm), which are Teflon-coated (DuPont, FEP, TE 9568).The use of stainless steel as cavity material is considered superior to glass, as used in our previous set-ups (Schuster et al., 2009;Thieser et al., 2016) as it eliminates the risk of breakage, and provides more homogeneous heating of the cavities.Radical losses on FEP-coated stainless-steel or glass surfaces are indistinguishable.PFA T-pieces (Swagelok) are mounted at both ends of the piping for admitting and exhausting the sample flow.Metal bellows/adjustable mirror supports are positioned behind each T-piece, resulting in a distance between mirrors of 93 cm.The bellows provide a purge gas volume for keeping the mirrors clean and reducing physical stress on the mirror supports.The 10 cavity mirrors are continuously purged with dry, synthetic air to prevent degradation in reflectivity.Purging the mirrors has the drawback of reducing the effective cavity length, and thus sensitivity, but it enabled uninterrupted operation over periods of several months.The 10 different purge gas streams are controlled by two mass flow controllers coupled with two sets of crit-ical orifices designed for different purge flow rates depending on the size and direction of flow of ambient air through the cavities.The purge gas flows for the 405 nm cavities are 500 cm 3 (STP) min −1 (sccm) for the mirror opposing the main gas flow (downstream) and 100 sccm for the upstream mirror.For the 662 nm cavities, these flows are 350 (downstream) and 150 sccm (upstream).The light intensity exiting each cavity is measured and converted to an analogue electric signal by a photomultiplier/preamplifier device (Hamamatsu H10492-012 at 662 nm and H10492-002 at 405 nm).Temperature, pressure and flows are measured/regulated by a custom-designed control unit (referred to as V25).Data acquisition (ring downs and laser spectra) and data processing are performed by an embedded computer (NI PXIe-8135).Analogue-to-digital conversion of the photomultiplier signals is performed via two acquisition cards (NI PXI-6132, 14 bit, 3 MS s −1 ) in the embedded computer rack, one for the two 662 nm cavities and one for the three 405 nm cavities.
2.1 Detection of NO 3 and N 2 O 5 at 662 nm NO 3 and N 2 O 5 are detected using the strong absorption feature of NO 3 at about 662 nm (σ 662 nm ≈ 2.3 × 10 −17 cm 2 molecule −1 at 298 K; Orphal et al., 2003;Osthoff et al., 2007).NO 3 is measured in one channel (the NO 3 channel) at room temperature.In a second channel (the N 2 O 5 channel), the sum of ambient NO 3 and NO 3 arising from the thermal decomposition of N 2 O 5 at 383 K are measured, enabling the N 2 O 5 mixing ratio to be calculated from the difference.The cavity mirrors (Advanced Thin Films, 1 in.diameter, 1 m radius of curvature) have a nominal reflectivity of 0.999985 and result in ring-down times (at atmospheric pressure of dry synthetic air) of 150-160 µs.To keep a constant emission wavelength of ≈ 662 nm, the laser diode (Thorlabs HL6545MG) is thermostatted to 38 • C and has an emission bandwidth of ≈ 0.5 nm (full width half maximum).An optical isolator is positioned at the immediate output of the diode laser to prevent back reflexions, and the final intensity of light reaching the front mirror of each of the two red cavities is ≈ 10-15 mW.The laser wavelength is monitored using a mini spectrograph (Ocean Optics, type HR4C2509), which records diffuse back reflexions from the N 2 O 5 cavity input mirror.Taking into account the emission spectrum and the absorption cross section of NO 3 , the effective cross section obtained during the first field deployment and calibrations is ≈ 2.1×10 −17 cm 2 molecule −1 .The uncertainty on the effective cross section is discussed in Sect.3.1.
Before entering the NO 3 and N 2 O 5 channels, air (≈ 15 L (STP) min −1 , slm) is drawn through a filter (Pall Corp., Teflon membrane, 47 mm ∅, 0.2 µm pore) housed in an automatic filter changer.The changer has a maximum capacity of 20-25 filters, allowing for 2-3 nights of measurement (at a change rate of one filter per hour) without requiring the presence of an operator.Losses of NO 3 in the filter and filter changer are discussed in Sect.3.1.Air exits the fil-ter changer through a 1/4 PFA pipe (≈ 30 cm) with a T-piece to enable the injection of a few standard cubic centimetres per minute of NO (100 ppmv in N 2 ) for titrating NO 3 and zeroing the instrument, which was performed for a few seconds every 4 min.The air then passes through another T-piece that divides the flow between the NO 3 (8 slm) and N 2 O 5 (7 slm) channels.For N 2 O 5 detection, the air flows through a heated Teflon-coated glass section (wall temperature 383 K) to decompose N 2 O 5 into NO 3 + NO 2 .This temperature is sufficient to decompose all the N 2 O 5 , which was verified experimentally by increasing the temperature until the maximum signal was obtained.The residence time in this reactor (≈ 0.2 s) is sufficient to allow for a reaction between total NO 3 and NO to go to completion during zeroing periods.In the NO 3 channel, the air flows first though FEP-coated glass tubing at room temperature to ensure sufficient time for complete NO 3 titration when adding NO (i.e. when zeroing).Residence times are 0.45 s in the NO 3 channel and 0.38 s in the N 2 O 5 channel.The pressure in the NO 3 and the N 2 O 5 channels is not regulated, and it depends on atmospheric pressure and flow rates in the instrument.
A square-wave signal provided by the V25 is used to modulate the diode laser on and off, and also triggers the acquisition of the individual decays.Typically, the modulation frequency is 400 Hz, with on and off times of 1 and 1.5 ms.Averaging 400 decay constants thus resulted in an instrumental time resolution of one data point per second.An Allan deviation plot obtained under laboratory conditions (800 mbar cavity pressure) for the NO 3 and N 2 O 5 cavities is shown in Fig. 3.The Allan deviation plot shows that the (1σ ) standard deviations of the zero signal (obtained in this experiment by sampling zero air) for the NO 3 and N 2 O 5 channels for a 1 s integration time are 0.1 and 2 pptv, respectively.This is found to be a significant improvement in signal-to-noise ratio compared to the instrument described by Schuster et al. (2009), for which 5 s integration times were necessary to reduce noise levels to ≈ 2 pptv for both NO 3 and N 2 O 5 channels.The improvement is likely related to the more stable optical set-up.Despite the fact that the two 662 nm channels in the new instrument are mechanically and optically very similar, the N 2 O 5 channel shows a higher noise level, which is related to enhanced turbulence due to temperature and thus density gradients.

Detection of NO 2 , PNs and ANs at 405 nm
Three channels sampling in parallel are used for the detection of NO 2 and organic nitrates.The three cavities are maintained at slightly above ambient temperature (305 K) to reduce density fluctuations.All three channels detect NO 2 (ambient, or produced by organic nitrate decomposition) using the absorption of NO 2 at 405 nm (σ 405nm at 298 K ≈ 6 × 10 −19 cm 2 molecule −1 ; Voigt et al., 2002).The cavity mirrors (Advanced Thin Films) have a reflectivity R ≈ 0.99995, which results in an average ring down of 30-35 µs at 700 mbar of dry synthetic air.The 405 nm cavities have a data acquisition system and optical set-up that is similar to the 662 nm cavities, although an optical isolator to prevent back reflections was found by empirical observation to be unnecessary at 405 nm.To keep the peak emission wavelength at 405 nm, the laser diode is held at 35 • C. The maximum intensity coupled into each of the three 405 nm cavities is in the range 10-15 mW.The laser wavelength is monitored using a mini spectrograph (Ocean Optics, type HR4C5451), which records diffuse back reflexions from the ANs cavity input mirror.The effective cross section for the three 405 nm channels during the calibrations and field deployment was 5.9 × 10 −19 cm 2 molecule −1 (based on the laser emission spectrum and the NO 2 absorption cross section).
Ambient NO 2 is detected directly when sampling from a room temperature inlet, whereas PNs and ANs are thermally decomposed in two heated sections of glass tubing at ≈ 448 and ≈ 648 K and then detected as NO 2 .In the 448 K channel ( PNs channel), the sum of ambient NO 2 plus NO 2 from the thermal decomposition of PNs is measured.In the 648 K channel ( ANs channel), NO 2 from the decomposition of ANs is additionally detected.The temperature set in the PNs channel (448 K) was chosen by measuring the temperature dependence of the thermal dissociation of PAN (≈ 2 ppbv) in synthetic air as provided by a diffusion source of PAN in tridecane.The normalised [NO 2 ] signal as a function of temperature (Fig. 4) shows that complete decomposition to NO 2 is achieved at a nominal temperature of ≈ 433 K.The temperatures given in the text and in Fig. 4 correspond to the external surface of the heated inlets and not necessarily that of the gas flowing through them.PANs have a thermal stability that is largely independent of the R group (Roberts, 1990;Kirchner et al., 1999).For the detection of atmospheric PNs, we set the temperature to 15 K above the threshold measured in the laboratory for PAN, i.e. 448 K. Figure 4 also shows the temperature-dependent decomposition of a sample of isopropyl nitrate in synthetic air: a yield of 100 % is reached at a set temperature of ≈ 633 K.As for PANs, the thermal decomposition of all ANs is expected to be similar, so a working temperature of 648 K was chosen.
Rayleigh scattering of 405 nm light by air (Fuchs et al., 2009;Thieser et al., 2016) results in sensitivity of the ring-down constant to pressure changes, and requires that the three 405 nm cavities are actively pressure-regulated.For this reason, and because zeroing requires overfilling the inlet line (see below), the 405 nm channels have an independent inlet system.In total, ≈ 8 slm air is sampled through 1/4 in.PFA piping, of which 0.5 slm is directed to a digital humidity sensor.Active pressure control of the cavities is achieved by pumping a regulated fraction of the main flow (usually ≈ 0.5 slm) through a flow controller directly to the exhaust system (see Fig. 2) to maintain the desired pressure (generally about 700 mbar).About 7 slm of ambient air is drawn through a filter held by a PFA filter holder.The particle filters used to protect the 405 nm cavities are the same as for the 662 nm cavities, but are changed at a much lower frequency (once per day per hand) due the lower reactivity of NO 2 and organic nitrates to surfaces.After passing through the particle filter, air is then split into three equal flows (≈ 2.3 slm), resulting in residence times of 1.7, 1.3 and 1.1 s for the NO 2 , PNs and ANs channels, respectively.The three 405 nm channels consist of vertically mounted glass tubing (length ≈ 55 cm, ID = 12 mm) connected to the cavities by PFA fittings as described in Sect. 2. For the PNs and ANs channels, a portion of the glass tubing is wrapped with heating wire and heavily insulated to achieve the desired thermal dissociation temperatures of 448 and 648 K.The first 10 cm of the heated section of the glass tubing in the PNs and ANs channels is filled with glass beads (≈ 0.5 mm diameter, Sigma-Aldrich G9268) supported on a glass frit (ca. 2 cm long).More detail and a diagram is given in the Supplement.The surfaces of the glass beads and glass frit scavenge organic radicals formed in the thermal decomposition of PNs and ANs, and thus reduce the impact of radical recombination with NO 2 and oxidation of NO, which can bias the results obtained, especially under conditions of high NO x (Day et al., 2002;Paul and Osthoff, 2010;Thieser et al., 2016).Details of the laboratory experiments carried out and the corrections necessary to take such processes into account are described in Sect.3.2.
The three 405 nm channels are zeroed for 30 s every 4 min by overflowing the inlet with ≈ 0.2 slm of bottled synthetic air or with scrubbed air provided by a zero-air generator (CAP 180, Fuhr GmbH).An upper limit to the NO 2 content of the zero air of ≈ 20 pptv was obtained by use of a blue-light converter, as described previously (Thieser et al., 2016).Previous work has shown that the different scattering coefficients for humid, ambient air and the dry, synthetic air used to zero the instrument can lead to a bias in the measured [NO 2 ].The difference in the Rayleigh scattering cross section at wavelengths close to 405 nm between synthetic air and water ( σ ) is reported to be between 0.4 and 0.5 × 10 −21 cm 2 molecule −1 (Fuchs et al., 2009;Thieser et al., 2016).At 50-60 % humidity at 30 • C and 1 bar, this results in an offset of ≈ 200 pptv in the [NO 2 ] signal.Thieser et al. (2016) have indicated that the 20 % difference in the reported scattering cross sections renders this a significant source of uncertainty at high relative humidity and low NO 2 .To avoid introducing additional error by correcting for this, we implemented a zero-air humidification system, which enables the 7 slm air flow used for zeroing the instrument to be actively matched (to better than 2 % up to 80 % relative humidity; RH) to ambient RH (of ambient and zero-air, both measured at 0.2 Hz).This was achieved by passing a variable fraction of the zero air to a bubbler containing distilled water.Taking σ = 0.5 × 10 −21 cm 2 molecule −1 , the error arising for a 2 % humidity difference at room temperature and 1 bar (measuring conditions of both humidity sensors) is equal to 10 pptv.
As the ring-down times are shorter at 405 nm (typically 35 µs), the diode laser modulation frequency could be increased so that 1666 ring-down constants were obtained per second.Figure 3 shows an Allan deviation plot (1σ ) for the NO 2 , PNs and ANs channels.The 1σ standard deviations for a 1 s integration time for the three 405 nm channels are 25, 40 and 45 pptv, respectively.The higher noise level associated with the PNs and ANs channels results from slightly lower mirror reflectivity (different batch) compared to the NO 2 cavity.

Data corrections, precision and total uncertainty
When based on known absorption cross sections, CRDS is an absolute concentration measurement technique.However, a number of corrections are necessary to convert measured ring-down times into volume mixing ratios of absorbing trace gases.Some corrections are related to the physical construction of the cavities, and others are related to chemical processes in the inlets and cavities that can bias the concentrations measured.The size and accuracy of these corrections contribute significantly to the overall uncertainty of the measurements, and are discussed in detail below.The more straightforward corrections needed are related to the effective absorption path length (l-to-d ratio), losses of NO 3 radicals in inlets and filters as previously described (Schuster et al., 2009;Thieser et al., 2016) and formation of NO 2 during the sampling time through O 3 + NO reaction.For this instrument, the effective l-to-d ratio is 0.69±0.02for the channels detecting at 405 nm, and 0.77 ± 0.04 for the channels detecting at 662 nm.Based on several laboratory experiments, the transmission of NO 3 in the filter/filter holder is 70±3 %.The 3 % uncertainty was obtained by repeated measurements under laboratory conditions.During the NOTOMO campaign (see Sect. 4), no discontinuities in the NO 3 signal were observed after hourly filter changes, which implies that there was no measurable change in transmission over the hour of exposure.In highly polluted environments, or those with highly reactive aerosol, this may not be the case, and more frequent filter changes may be necessary to avoid loss of NO 3 .No losses of N 2 O 5 on passage through the filter holder are observed under laboratory conditions, and similarly to NO 3 , no evidence was obtained for losses on aged filters during the NOTOMO campaign.The transmission of NO 3 in the NO 3 and N 2 O 5 channels is 97.4 ± 2.5 % and 90.2 ± 5 %, respectively.Details and results of the experiments to derive l-to-d ratios and correction factors for NO 3 losses are given in the Supplement.Corrections related to NO oxidation by O 3 are described in Sect.3.2.5.

NO 3 and N 2 O 5 channels
For averaging periods of longer than ≈ 1 s, the detection limit of both [NO 3 ] and [N 2 O 5 ] is defined mainly by variation/drift in the zero signal.Drifts can arise from changes in the mirror reflectivity and through thermal and mechanical stresses that influence the cavity alignment.The detection limit can be estimated by the 2σ standard deviation of the difference from one zeroing period to the next one for the whole campaign.The values for the NO 3 and N 2 O 5 channels are respectively 1.5 and 3 pptv, and imply that 95 % of the time, the difference between one zero value to the next one is lower than 1.5 and 3 pptv respectively.The NO 3 limit of detection is thus 1.5 pptv for 1 s averaging and zeroing every 3 min.As N 2 O 5 is obtained by difference, the detection limit is ≈ 3.5 pptv (also for 1 s averaging).The total uncertainty for the [NO 3 ] measurement can be estimated by propagation of the individual uncertainties associated with the correction factors (see above) and the uncertainty associated with the effective absorption cross section, which depends on the NO 3 absorption cross section uncertainty (≈ 10 %, taken from Osthoff et al., 2007) and the uncertainty associated with measurement of the laser emission spectrum (≈ 5 %).The resulting total uncertainty for NO 3 measurement is 25 %.The uncertainty associated with the [N 2 O 5 ] measurement depends on the absolute values of [NO 3 ] and [N 2 O 5 ], and is therefore variable.As an example, combining the 25 % uncertainty of the total NO 3 measured in the N 2 O 5 channel with [NO 3 ] and [N 2 O 5 ] mixing ratios of 50 and 500 pptv results in an uncertainty of 28 % for the N 2 O 5 measurement.The transmission of NO 2 through the hot glass inlets (with frit and packed with glass beads) of the PNs and ANs channels was investigated by sampling NO 2 in synthetic air (up to 10 ppbv) into the three 405 nm channels simultaneously.NO 2 is depleted slightly in the hot inlets with a transmission of 98.5±2 % and 95.5±2 % found for the PNs and ANs inlets, respectively (see Fig. 5).The NO 2 transmission was measured before, during and after the field deployment of this instrument (corresponding to a period of several months), and was found to be constant and also independent of relative humidity.

Reaction of radicals with NO / NO 2 in the PNs channel (448 K)
The thermal decomposition of PNs leads to the formation of organic radicals and NO 2 .In an ideal situation, in which PNs decompose 100 % to NO 2 and when NO 2 does not undergo any further production or loss reaction, the total [NO 2 ] measured in the PNs channel is equal to the sum of ambient NO 2 and PNs.Under the operating conditions of the CRDS instrument described here, the total [NO 2 ] signal in the PNs channel can, however, be biased by a number of reactions initiated by the organic radicals.The influence of these reactions has been described in the literature (Day et al., 2002;Paul and Osthoff, 2010;Thieser et al., 2016).Taking the example of peroxyacetyl nitrate (PAN), subsequent to its thermal decomposition (Reaction R1a), there are processes that lead both to the removal of NO 2 (Reactions R1b, R5) and to its formation (Reactions R2, R3, R4).
Normalised signal Reactions (R2)-(R4) show that, in the most unfavourable scenario, if all organic radicals react with NO, each peroxyacetyl radical can lead to the formation of three NO 2 molecules, biasing the result by the same factor.Likewise, the presence of very high NO 2 could conceivably result in complete reformation of PAN (Reaction R1b).The size and sign of the bias thus depends on the relative concentrations of NO and NO 2 and, most importantly, on the rate of loss of organic radicals to the reactor walls or via thermal decomposition.The bias resulting from these reactions can be reduced by minimising the residence time between thermal dissociation and detection and by making pressure-dependent recombination reactions inefficient by working at low pressures, e.g. by using LIF detection of NO 2 (Day et al., 2002;Wooldridge et al., 2010).As CRDS instruments operate at higher pressures, to maintain sufficient sensitivity (typically from 0.5 to 1 bar), we have taken a different approach and optimised the surface losses of the organic radicals by modifying the surface-to-volume ratio in the heated inlets.First experiments on PAN samples using glass wool as a radical scavenger resulted in the desired reduction in the rate of recombination of CH 3 C(O)O 2 with NO 2 .Glass wool was, however, observed to greatly enhance (rather than reduce) the oxidation of NO to NO 2 .This is presumably the result of a surface-catalysed process as previously observed on powdered, aluminium silicate mineral dust samples (Hanisch and Crowley, 2003), and may be related to the formation of ox-idised surface sites that can react with NO.This led us to test glass beads, which presumably have less reactive "defective" sites than glass wool but a sufficient surface area to remove a large fraction of the organic radicals (see below).We conducted a series of experiments with different mixtures of PAN and NO 2 / NO (from 1 ppmv in synthetic air gas bottles) and analysis of the observations by numerical simulation, similar to that presented in Thieser et al. (2016).
The set of chemical reactions used in the numerical simulations is essentially the same as that described in Thieser et al. (2016).One exception is formation of NO 2 via the reaction between O 3 and NO, which Thieser et al. (2016) treated as occurring independently of other chemical processes.Here, this reaction is treated more rigorously by including it in the chemical simulation.The other difference is the use of two different values for a rate constant (k s ) used to calculate the kinetic limitation on the radical uptake coefficient (γ , see below).In order to simulate all laboratory data sets, different values of k s (one for sections A and B and one for sections C and D; see Fig. S1 in Supplement) are required to account for the temperature gradient in the heated sections of the inlets.The heterogeneous loss of radicals to the reactor walls is based on a Langmuir-Hinshelwood mechanism, with the first-order rate constant (k w ) given by Eq. ( 2).
The uptake coefficient, γ , depends on both kinetic and diffusive limitations as described in detail by Thieser et al. (2016).
Here we focus on the role of glass beads in enhancing the rate of uptake and reaction of organic radicals by increasing the surface area available for reaction (A).Without the glass beads, the value of A in the heated inlets is 3.5 cm 2 cm −3 (corresponding to a cylinder with internal diameter 1.2 cm).The presence of the glass beads, with an average diameter of 0.5 mm, results in a surface-to-volume ratio of ≈ 100 cm 2 cm −3 .The exact determination of the temperature profile in the inlet is possible in the case of a classic cylindrical reactor, but is difficult to achieve with the presence of the glass beads and glass frit.An approximate profile in section D of the reactor (part of the cylindrical reactor between the fritted glass and the cavity inlet; see Fig. S1) was obtained by measuring the temperature between the fritted glass and the entrance of the cavity.It is assumed that the temperature in section B (containing the glass beads) increases from ambient to 448 K at the point where the gas enters section C (corresponding to the fritted glass section).This profile is also used to calculate the residence time in the PNs channel of 1.3 s.To take into account the variations of temperature, pressure, specific surface area and diffusion radius along the glass reactor, a profile for each of these parameters is used as input in the numerical simulation.
Figure 6a shows results of a set of four experiments in which different amounts of PAN are mixed with different amounts of NO 2 and monitored in the three channels.
The y axis shows the total [NO 2 ] measured in the PNs channel ([NO 2 ] 448 K ) minus NO 2 measured in NO 2 channel ([NO 2 ] Amb ).The negative slopes of these data indicate that NO 2 from PAN decomposition is lost at high mixing ratios of added NO 2 .A similar set of data, but with added NO rather than NO 2 , is displayed in Fig. 6b.In this case, the positive slopes indicate that NO 2 is being formed from the reaction of organic radicals with NO.Note that the solid lines are the results of the numerical simulations (with identical mechanism) of NO 2 formation and loss in both experiments.The mixing ratios of PAN listed are those required in the model to match the experimental data.In both data sets the slopes are weakest at low PAN concentrations, reflecting the fact that radical-NO or radical-NO 2 reaction rates will be dependent on the radical concentrations.The mixing ratios of PAN cannot be derived by simple back extrapolation to zero NO 2 or NO as this does not take into account the NO 2 formed when PAN itself decomposes.At the highest mixing ratios of NO 2 , up to 25 % of the peroxyacetyl radical recombines with NO 2 while the other 75 % is lost to the walls or did not react with NO 2 .
The analysis of the measurements shows that (without correction) NO 2 would be overestimated by approximately 40 % for 3 ppb of PAN when sampling air containing 8 ppbv of NO.This is much lower that the ≈ 150 % reported by Thieser et al. (2016) for similar conditions in their PNs channel, reflecting the increased loss of radicals on the glass surfaces.
In both series of experiments, the model reproduces the data for the whole range of [PAN], [NO 2 ] and [NO] explored.A set of experiments was carried out to investigate whether the formation (or loss) of NO 2 described above could be influenced by the presence of water vapour acting, for example, as a quencher of a surface reaction.A mixture of PAN plus NO 2 was humidified to 60 % RH, with other operating conditions similar to the previous experiments.The numerical model reproduced the measurement data without the need to modify the wall loss rates or add any chemical processes involving H 2 O, suggesting that the gas and surface reactions taking place are not significantly influenced by adsorbed water at these temperatures.In order to correct field data for these biases, an iterative fitting procedure is used and described in detail in Sect.4.2.

Reaction of radicals with NO / NO 2 in the ANs channel (648 K)
As first discussed by Thieser al. (2016), experimental and theoretical studies show that the higher temperature (648 K) of the ANs channel changes the chemical processes substantially compared to 448 K. ANs can now decompose to NO 2 and an alkyl radical fragment, whereas the peroxyacetyl radical (from PAN decomposition) is thermally unstable.The NO 2 generated in the ANs channel when adding NO 2 / NO to PAN samples is displayed in Fig. 6c and d  The model takes into account two main pathways for the fate of the peroxyacetyl radical (Reactions R6, R7) in which the acetyl radical (CH 3 CO) formed in Reaction (R6) can react further with O 2 to reform a peroxyacetyl radical or an OH radical and an α-lactone (Reactions R8, R9) (Carr et al., 2011).CH 2 C(O)OOH can also decompose to OH and an α-lactone (Reaction R10) (Carr et al., 2007(Carr et al., , 2011;;Chen and Lee, 2010).If not lost to the walls, OH can react with NO 2 to form HNO 3 according to Reaction (R5).
The acetyl radical can also form a methylperoxy radical CH 3 O 2 through the formation of a CH 3 radical and CO.The net reaction is given by Reaction (R11).In the presence of NO, the peroxy radicals (CH 3 C(O)O 2 , CH 3 O 2 and HO 2 ), if not lost to the glass surface, can lead to the formation of NO 2 .
To simulate the reactions the ANs channel, the temperature profile was determined in the same way as the PNs channel profile.Due to the higher gas flow velocity at the elevated temperatures, the pressure profile is slightly different than in the PNs channel and gives a cavity pressure that is about 10 mbar lower.The results of the experiments involving addition of NO 2 to PAN samples in the ANs channel are presented in Fig. 6c.For the range of PAN concentrations covered, the loss of NO 2 by chemical recombination is about 5 % of the initial PAN.The initial PAN concentration used as input variable for the numerical simulations is, for all four experiments, about 3-5 % higher than that needed to fit the data obtained in the PNs channel (Fig. 6a), which is most probably related to model uncertainties.Since this 3-5 % difference is the same for all four experiments and is apparently not correlated with the amount of PAN, it is added to the calculation of total uncertainty for the PNs measurement.The results of the NO addition experiments are shown in Fig. 6d.The thermal decomposition of the peroxyacetyl radical reduces the positive bias due to NO oxidation so that a maximum factor of only 1.04 (compared to 1.40 in the PNs channel) is obtained.The initial PAN concentration required to fit the NO addition data set at 648 K is 5-10 % higher than at 448 K, which is related to uncertainty in the [NO] mixing ratio used or a bias in the simulation.In order to investigate the role of organic (alkyl) radicals generated from the thermal decomposition of ANs, a set of experiments was conducted in which a sample of i-propyl nitrate (C 3 H 7 ONO 2 ) in synthetic air was mixed with different amounts of NO and NO 2 .Thermal decomposition of i-propyl nitrate (Reaction R12) is followed by Reactions (R13) and (R14) that generate HO 2 and CH 3 O 2 , which can convert NO to NO 2 and sequester NO 2 as HO 2 NO 2 and CH 3 O 2 NO 2 .
Figure 7a and b show that the addition of NO increases the amount of NO 2 formed per AN, whereas the presence of NO 2 results in a negative bias to the data.These effects are captured well for both data sets by the model simulations (blue lines).

Effect of thermal decomposition of O 3
At high temperatures, O 3 decomposes according to Reaction (R15).Although most of the O atom produced reacts with O 2 to reform ozone (Reaction R16), it can form NO+O 2 in the presence of NO 2 (Reaction R17).
The importance of this process is strongly dependent on the operating conditions of the instrument, recent studies showing that higher temperatures or pressures may result in a significant, negative bias from the ozone-pyrolysis-initiated reduction of NO 2 (Lee et al., 2014;Thieser et al., 2016).Assuming that the steady-state concentration of the O atom in the heated inlets is dominated by Reactions (R15) and (R16) ), the loss of NO 2 can be approximated as follows: The term in brackets is constant at constant temperature, pressure and flow rate through the heated inlet, and can be determined experimentally by sampling different mixtures of NO 2 and O 3 simultaneously through the ambient temperature channel and the two hot channels.Figure 8 shows the difference [NO 2 ] Amb − [NO 2 ] 648 K (in ppbv) as a function of [NO 2 ] × [O 3 ] for three different levels of ozone (50, 115 and 185 ppbv).A linear fit through the data points yields a slope of 2.54 ± 0.26 × 10 −4 ppbv −1 , which is < 50 % of the value found by Thieser et al. (2016).The difference can be attributed mainly to the use of a lower oven temperature (648 K in this work instead of 723 K).No reduction of NO 2 is observed in the PNs channel because thermal decomposition of O 3 at 448 K is too slow.To illustrate the impact of the thermal dissociation of O 3 in the ANs channel, we note that 5 ppbv of total NO 2 in the presence of 50 ppbv of O 3 results in the removal of 75 pptv NO 2 .Since the amount of O 3 decomposed and NO produced in this process is negligible compared to ambient amounts, it does not affect the input conditions for the numerical simulations, and can therefore be treated separately using Eq.(3).

Effect of NO oxidation by O 3
A positive bias to the measurement of NO 2 may result from NO oxidation by O 3 during the time it takes for the sampled air to flow through the inlets and cavities of the instrument (Reaction R18).
The rate coefficient of this reaction is temperaturedependent, and the effect is larger at 448 and 648 K than at room temperature.To investigate this interference, different mixtures of O 3 and NO were sampled simultaneously through the three channels.The results are shown in Fig. 9 where data from four experiments with different O 3 mixing ratios (41, 62, 80 and 131 ppbv) are plotted.The solid lines are fits to the data, the slopes of which are second-order rate constants multiplied by a reaction time (cm 3 molecule −1 ).The rate constant for the gas-phase reaction between NO and O 3 is listed as Atkinson et al., 2004).For the PNs channel (448 K), using this expression results in an underestimation of the effective (measured) production of NO 2 by a factor 1.06.For the ANs (648 K) channel the equivalent factor is 1.52.The most likely reason for this discrepancy is that some NO is oxidised in a surface catalysed process, presumably involving surface sites that are activated following interaction with O 3 .To account for this extra source of NO 2 , the O 3 plus NO reaction is implemented in the chemical model (with modified rate expressions) for the 448 and 648 K data corrections (see Sect. 3.2.2 and 3.2.3),and corrected manually for the NO 2 channel.

Detection of other species as NO 2 after thermal dissociation
Atmospheric trace gases other than PNs and ANs can potentially be detected as NO 2 in this instrument.As thermal decomposition of N 2 O 5 to NO 3 and NO 2 has been shown to be 100 % efficient in our 383 K inlet in the 662 nm channels, we expect 100 % dissociation at the higher temperatures of the PNs and ANs channels.This may represent a significant source of bias during the night when N 2 O 5 mixing ratios can be large.However, as the five-channel instrument simultaneously monitors N 2 O 5 , this can be easily corrected.Here we consider possible HNO 3 and ClNO 2 interferences in the ANs channel.HNO 3 is a major reservoir of tropospheric NO x .Although HNO 3 is thermally stable at temperatures below 700 K, Wild et al. (2014) report ≈ 95 % conversion of HNO 3 to NO 2 at 648 K, whereas Thieser et al. (2016) found about 10 % at 723 K.In order to test for unwanted detection of HNO 3 when sampling from the hot inlets (with glass beads) in the fivechannel instrument, we injected a sample of HNO 3 in synthetic air directly prior to the flow division before entering the three 405 nm channels.In this way, possible losses of HNO 3 in the inlet in front of the heated sections of the instrument were minimised.HNO 3 was generated in a custombuilt permeation source, which was calibrated by absorption spectroscopy at 184.85 nm, where HNO 3 absorbs strongly.The permeation source generated several ppmv of HNO 3 in a flow of 10 sccm, which, following dilution, provided a mixing ratio of ≈ 30 ppbv at the CRDS inlet.The permeation source also generated about 3 ppbv of NO 2 .Figure 4 (open circles) shows that NO 2 arising from HNO 3 decomposition could not be observed in either heated inlet, enabling us to set an upper limit to the decomposition efficiency of HNO 3 to NO 2 of < 0.5 %.This is in broad agreement with the previous measurement by Thieser et al. (2016), who saw no evidence for HNO 3 decomposition below 650 K, and is in stark contrast to that reported by Wild et al. (2014).While we have no rigorous explanation of this, we note that the gas-phase thermal dissociation of HNO 3 to NO 2 at the temperature of these experiments is too slow to explain its formation, which suggests that it is probably surface-catalysed in the experiments of Wild et al. (2014), implying that different glass types or chemical history of the surface may influence the formation of NO 2 significantly.
As reported previously, ClNO 2 , formed in the atmosphere in the reaction between N 2 O 5 and chloride-containing particles, can be detected as NO 2 in TD-CRDS instruments (Thaler et al., 2011;Wild et al., 2014;Thieser et al., 2016) The ClNO 2 decomposition efficiency as a function of the oven temperature is plotted in Fig. 4. The ClNO 2 sample was generated by flowing Cl 2 in synthetic air over sodium nitrite crystals.In normal operating conditions (at 648 K), 90 ± 3 % of the ClNO 2 is decomposed to NO 2 .As ClNO 2 mixing ratios are highly variable and can approach ppbv levels, ClNO 2 may thus represent a serious limitation to ANs measurements, unless independent measurements are available to enable correction.
3.2.7 Detection limit and total uncertainty for NO 2 , PNs and ANs The detection limit for the 405 nm channels can be estimated in a similar manner to that described for the 662 nm channels.The total uncertainty associated with the [NO 2 ] measurement is a combination of the uncertainties in the l-to-d ratio, humidity matching of the zero and ambient air and the correction for the NO + O 3 reaction on the absorption cross section of NO 2 and on N 2 O 5 decomposition.The total uncertainty on the absorption cross section is estimated to 6 %, taking into account the error on the reference cross section (Voigt et al., 2002) and fluctuation in the laser emission spectrum.This value, when combined with the uncertainty associated with the l-to-d ratio, results in 6.5 % uncertainty.The uncertainty associated with the correction for NO 2 formation in the NO + O 3 reaction depends on ambient ozone and NO levels.Considering 10 % uncertainty for k O 3 +NO and 5 % uncertainty for [O 3 ], [NO] and [NO 2 ], [NO] and [O 3 ] ambient levels of 1, 0.5 and 50 ppbv, we obtain an uncertainty for this correction of 0.1 % on the final [NO 2 ] value.NO 2 can be formed at room temperature by the slow thermal decomposition of N 2 O 5 .Taking a decomposition rate constant of 4.4×10 −2 s −1 at 303 K for N 2 O 5 (Atkinson et al., 2016) and a residence time in the NO 2 channel of 1.7 s, we calculate that 0.08 of the N 2 O 5 can decompose.The largest N 2 O 5 -to-NO ratio measured during NOTOMO was 0.17, which results in a maximum contribution of ≈ 1.5 % NO 2 from N 2 O 5 decomposition.As mentioned in Sect.2.2, the upper limit for bias caused by humidity matching errors when zeroing is 10 pptv, which results in a final uncertainty for [NO 2 ] of 8 % + 10 pptv.
The uncertainty of the total NO 2 detected in the PNs and ANs channels is a combination of the same uncertainties involved in the [NO 2 ] measurement plus the uncertainty of the NO 2 transmission through the heated glass beads.Taking the values listed in Sect.3.2.1,we obtain an uncertainty for total NO 2 detected of 7 % pptv for both PNs and ANs channels.To correct for the radical plus NO x biases, the to-tal NO 2 signals measured in the two hot channels are used to constrain the chemical model described earlier.The uncertainty arising from these calculations depends on the total [NO 2 ] measured in each channel, and is thus highly variable.Note that uncertainty arising from the humidity difference is cancelled out by the subtraction to obtain [PNs] and [ANs].The uncertainty added by the calculation of the ambient [PNs] and [ANs] depends on the model uncertainties, as well as on the uncertainty associated with the trace gas concentrations involved in this calculation.The uncertainty associated with [ PNs] and [ ANs] is related to ambient [NO 2 ], [O 3 ] and [NO].The fact that the corrections rely on PAN and IPN chemistry led Thieser et al. (2016) to estimate the uncertainty associated with the model-derived corrections to be 30 % (max) of the overall correction factor.The final uncertainties on the [ PNs] and [ ANs] value depend on the total NO 2 signal in all three channels, and can vary significantly.Below, we discuss the correction factors required for the NO 2 , PNs and ANs data sets obtained during the first field deployment of the instrument.

First field deployment of the five-channel CRDS
The instrument described here was first deployed during the NOTOMO campaign (Nocturnal Observations at the Taunus Observatory; insights into Mechanisms of Oxidation) that took place during summer 2015.The site, previously described in detail (Crowley et al., 2010;Sobanski et al., 2016), is situated on top of the Kleiner Feldberg mountain (≈ 850 m a.s.l.) at the southern limit of the forested Taunus mountain range and north of the Frankfurt-Mainz-Wiesbaden agglomeration.The set of instruments deployed during NOTOMO was located in two research containers, and sampled from a common inlet.Approximately ≈ 10 m 3 min −1 air was drawn from a height of 8 m above ground through a 15 cm diameter stainless-steel pipe connected to an industrial fan.Gases were sampled (≈ 23 slm) in the TD-CRDS (located in the upper container) from the centre of the high-flow inlet via 1/2 in.PFA tubing (for the two 662 nm channels) and 1/3 in.PFA tubing (for the three 405 nm channels) ≈ 3.5 m from the top of the inlet.Residence times in the high flow inlet, in the 1/2 in.PFA tubing and in the 1/4 in.PFA tubing were 0.3, 0.05 and 0.1 s, respectively.In the following sections we shall not attempt to perform a detailed analysis of the entire campaign data set, which is beyond the scope of this paper, but provide a more qualitative description of the results, and indicate the size of corrections applied and the total uncertainty.with an average value of 1.8 ppbv.NO 3 and N 2 O 5 mixing ratios reach up to 40 and 400 pptv at night, and are below the detection limit during the day.These values are in the range of previous measurements at this location (Crowley et al., 2010;Sobanski et al., 2016), in which NO 2 , NO 3 and N 2 O 5 mixing ratios of up to 20, ≈ 200 and ≈ 3000 pptv, respectively were measured.Within a few minutes after sunset, NO 2 , NO 3 and N 2 O 5 acquire thermal equilibrium and the equilibrium constant, K eq , can be calculated from [N 2 O 5 ] / [NO 2 ][NO 3 ].In Fig. 11 we compare temperaturedependent values of the equilibrium constant using measured concentrations of NO 2 , NO 3 and N 2 O 5 as well as values from the literature that are based on laboratory measurements (Atkinson et al., 2004;Burkholder et al., 2016).Within combined uncertainty, the values agree, though we note that the parameterisation of Burkholder et al. (2016) results in val- ues that are closer to those obtained by analysing the field measurements and that the agreement is better at lower temperatures.We are wary of over-interpretation of this fact and are aware that the laboratory determinations that led to the recommended values are expected to be more accurate, especially close to room temperature where many data sets exist and are in good agreement.In this context we note that small errors in the measurement of the ambient temperature or a 10-20 % inlet loss of NO 3 would have been difficult to observe but would have had a significant impact on the equilibrium constant calculated from the field data and may have contributed to the differences observed.

Mixing ratios of PNs and ANs
By simply subtracting [NO 2 ] measured by the NO 2 channel from the total NO 2 signal measured by the PAN channel and the total NO 2 signal in the PNs channel from the total NO 2 signal in ANs channel, we obtain uncorrected mixing ratios for PNs and ANs, which are plotted (black data points) in Fig. 10.In this context, uncorrected means that no chemical corrections have been applied (e.g. for reformation of PNs in the presence of NO 2 , or formation of NO 2 from NO oxidation with O 3 see Sect. 3) but physical corrections (e.g.l-to-d ratio) have been made.
The procedure used to correct the [ PNs] and [ ANs] mixing ratios for the effects described in Sect. 3 utilises an iterative algorithm based on the chemical model used to simulate the laboratory data described in Sect.3.2.To correct the PNs data, a simulation using [NO 2 ], [O 3 ] and [NO] values and an initial guess for [ PNs] were performed.Values of O 3 were taken from ambient measurements, whereas, in the absence of NO measurements, we assumed that its mixing ratio was zero at night-time and during the day could be calculated from its photochemical steady state, [NO] ss = J NO 2 [NO 2 ]/k NO+O 3 [O 3 ], where J NO 2 is the photolysis frequency of NO 2 (measured using a spectral radiometer) and k NO+O 3 is the rate constant for reaction of NO with O 3 .This method of estimating [NO] resulted in satisfactory agreement with measurements from the Hessian Agency for Nature Conservation, Environment and Geology (HLUG) for periods when NO was above the detection limit (> 1 ppb) of their instrument, which monitors NO permanently at the site.
In the iterative procedure, [ PNs] were tuned automatically until the simulated total [NO 2 ] matched the measured [NO 2 ] in the PNs channel.This procedure was applied to each data point (10 min resolution) obtained during the campaign.The corrected [ PNs] were used as an additional input to correct the [ ANs] data.In this case, constant values for [NO 2 ], [O 3 ], [NO 2 ] and [ PNs] and a variable value for [ ANs] are inputted into the model to match the total NO 2 signal measured in the ANs channel.For the correction of the [ PNs] data, ambient N 2 O 5 was taken into account by adding its mixing ratio to the initial [NO 2 ] values.This was also done also for the [ ANs] correction.Measured ClNO 2 mixing ratio was decreased by a factor of 0.9 to account for the decomposition efficiency and was then added to the initial [NO 2 ] for simulation of the total NO 2 in the ANs channel.The corrected data are displayed as red data points in Fig. 10a.
The overall correction factor (corrected data/uncorrected data) required for the PNs and ANs measurements are illustrated in Fig. 10b and c.The peak of the distributions is at ≈ 1.1 for both data sets, and the sigma values corresponding to a Gaussian fit are 0.06 and 0.18 for PNs and ANs respectively, indicating that the corrections required are dominated by NO 2 rather than NO, and that the corrections for the ANs vary more due to the presence of more organic radicals in the ANs channel.

Conclusions
We have constructed and characterised a five-channel, thermal-dissociation cavity ring-down spectrometer for measurement of the reactive nitrogen traces gases, NO 2 , NO 3 , N 2 O 5 peroxy nitrates and alkyl nitrates, which provides insight into the coupling of atmospheric, daytime and nighttime NO x and RO x chemistry.The total measurement uncertainties and limits of detection are estimated at 25 % and 2 pptv for NO 3 , 28 % and 10 pptv for N 2 O 5 and 6.5 % +10 and 59 pptv for NO 2 .
Chemical interferences in the measurements of PNs and ANs were reduced by the use of low TD temperatures and the use of glass beads as the radical scavenger.In combination with extensive laboratory tests to enable accurate cor-rection for bias in PNs and ANs measurements resulting from reactions of organic radicals formed in the thermaldissociation channels, this reduces overall uncertainty and extends the NO x regime in which the instrument can be operated.The total uncertainty associated with the PNs and ANs measurements is dependent on NO x levels and also on the presence (mainly at night) of ClNO 2 and N 2 O 5 .During the PARADE campaign the average correction was (1.1 ± 0.3) where the quoted uncertainty is an estimate of systematic error related to the use of a numerical model to simulate chemical processes in the TD sections of the instrument.

Data availability
The PARADE data can be obtained on request (via John Crowley) from the owners.The NOTOMO data will be released at the end of 2017.
The Supplement related to this article is available online at doi:10.5194/amt-9-5103-2016-supplement.

Figure 1 .
Figure 1.Atmospheric sources and sinks of the NO y species (in red circles) measured by the five-channel TD-CRDS instrument described in this work and their link to VOC degradation and photochemical O 3 formation.

Figure 2 .
Figure 2. Schematic of the five-channel TD-CRDS.Gas flows are represented by coloured arrows.The wall temperatures of the different inlets are given for each channel.RT: room temperature.

Figure 3 .
Figure 3. Allan deviation plot (1σ ) for each of the five CRDS channels.The data were obtained in the laboratory at ≈ 298 K and a cavity pressure of 800 mbar.

Figure 4 .
Figure 4. Decomposition profiles of PAN (red points), i-propyl nitrate (black points) and ClNO 2 (blue points) with statistical errors (1σ ).The PAN decomposition profile was measured in the PNs channel; the i-propyl nitrate (IPN) and ClNO 2 profile were measured in the ANs channel.The open black circles indicate the lack of HNO 3 decomposition when sampled into the ANs channel.
3.2 NO 2 , PNs and ANs channels 3.2.1 Losses of NO 2 in the heated PNs and ANs inlets

Figure 5 .
Figure 5. NO 2 loss in the heated inlets of the PNs and ANs channels.Black points represent the NO 2 mixing ratio measured in the PNs channel ([NO 2 ] 448 K ) normalised to the NO 2 mixing ratio measured in the NO 2 channel ([NO 2 ] Amb ).Red points represent the NO 2 mixing ratio measured in the ANs channel ([NO 2 ] 648 K ) normalised to the NO 2 signal measured in the NO 2 channel.

Figure 6 .
Figure 6.Modelled (lines) and measured difference between the NO 2 signal in the hot channels and ambient temperature channel for different PAN samples and different added amounts of NO and NO 2 .(a) Addition of NO 2 to the 448 K (PNs) channel.(b) Addition of NO to the 448 K (PNs) channel.(c) Addition of NO 2 to the 648 K (ANs) channel.(d) Addition of NO 2 to the 648 K (ANs) channel.

Figure 7 .
Figure 7. Modelled (lines) and measured difference between the NO 2 signal in the 648 K channels and ambient temperature channel for i-propyl nitrate (IPN) samples and different added amounts of NO 2 (a) or NO (b).

Figure 8 .
Figure 8. Influence of O 3 pyrolysis.Loss of NO 2 (in ppbv) in the heated ANs channel compared to the NO 2 channel vs. the product of the NO 2 and O 3 mixing ratios.

Figure 9 .
Figure 9. Production of NO 2 from the O 3 +NO reaction when sampling into the three 405 nm channels with different amounts of NO and O 3 .The numbers in parentheses are the slopes of least-squares fits (solid lines), which correspond to the product of the effective second-order rate constant and the reaction time.

4. 1
Figure10ashows the measured mixing ratios of NO 2 , NO 3 , N 2 O 5 , PNs and ANs for 1 week between 30 June and 7 July.During this period, [NO 2 ] varied from 0.3 to 6 ppbv,

Figure 10 .
Figure 10.(a) Mixing ratios of NO 2 , NO 3 , N 2 O 5 , PNs and ANs measured by the TD-CRDS instrument between 30 June and 7 July 2015 at the Kleiner Feldberg observatory, Germany, as part of the NOTOMO campaign.For the three lowest panels in ([NO 2 ], [NO 3 ] and [N 2 O 5 ], the black points represent the final corrected data.For the PNs panel, the black points correspond to uncorrected data obtained by subtracting [NO 2 ] measured in the NO 2 channel from the NO 2 measured in the PNs channel.The red points have been corrected using the chemical model with the iterative fitting procedure.For the ANs panel, the black points correspond to the uncorrected data (obtained by subtracting the NO 2 signal measured in the PNs channel from the NO 2 signal measured in the ANs channel).The red points have been corrected using the chemical model plus iterative fitting procedure.Panels (b) and (c) are frequency distributions for the correction factors applied using the iterative numerical simulations.

Figure 11 .
Figure11.Equilibrium constant (K eq ) for the reaction NO 2 + NO 3 = N 2 O 5 calculated from measurement of each trace gas (black points, with grey bars representing total uncertainty).The solid lines are values of K eq recommended by the JPL evaluation panel (red line;Burkholder et al., 2016) and IUPAC evaluation committee (blue line;Atkinson et al., 2004).The dashed lines represent the minimum and maximum values according to the evaluations.
The 2σ standard deviation for consecutive zeros for the NO 2 , PNs and ANs channels are respectively 59, 74 and 54 pptv.Using these values, we obtain detection limits for [NO 2 ], [PNs] and [ANs] of 59, 94 and 80 pptv.The detection limits for [PNs] and [ANs] are obtained by error propagation on the NO 2 and PNs channels and NO 2 and ANs channel (see Sect. 4).