Chemical ionization quadrupole mass spectrometer with an electrical discharge ion source for atmospheric trace gas measurement

We present a chemical ionization quadrupole mass spectrometer (CI-QMS) with a radio-frequency (RF) discharge ion source through N₂∕CH₃I as a source of primary ions. In addition to the expected detection of PAN, peracetic acid (PAA) and ClNO₂ through well-established ion–molecule reactions with I⁻ and its water cluster, the instrument is also sensitive to SO₂, HCl and acetic acid (CH₃C(O)OH) through additional ion chemistry unique to our ion source. We present ionization schemes for detection of SO₂, HCl and acetic acid along with illustrative datasets from three different field campaigns underlining the potential of the CI-QMS with an RF discharge ion source as an alternative to ²¹⁰Po. The additional sensitivity to SO₂ and HCl makes the CI-QMS suitable for investigating the role of sulfur and chlorine chemistry in the polluted marine and coastal boundary layer.

1 Introduction

Chemical ionization mass spectrometry using iodide anions (commonly referred to as I-CIMS) is a widely used technique to measure various atmospheric trace gases with a high temporal resolution and low detection limit. The potential of I-CIMS for atmospheric trace gas measurement was established in laboratory studies (Huey et al., 1995) on chlorine nitrate (ClONO₂), which plays a central role in polar stratospheric O₃ depletion (Molina et al., 1987), and dinitrogen pentoxide (N₂O₅), which, through its heterogeneous hydrolysis on cloud droplets and aerosols, acts as a sink of gas-phase NO_x (NO+NO₂) (Lelieveld and Crutzen, 1990; Dentener and Crutzen, 1993). The first applications of I-CIMS for monitoring atmospheric composition were for measurement of N₂O₅, peroxyacyl nitric anhydride (PAN, CH₃C(O)O₂NO₂) and other peroxycarboxylic nitric anhydrides (Huey, 2007). Since then it has been recognized that several classes of organic and inorganic trace gases can be detected sensitively by I-CIMS including organic and inorganic acids, organic nitrates, halogen nitrates and di-halogens (Phillips et al., 2013; Lee et al., 2014; Neuman et al., 2016; Priestley et al., 2018). Our instrument has previously been deployed with a radioactive ion source (²¹⁰Po) to investigate the atmospheric chemistry of nitryl chloride (ClNO₂), PAN and peracetic acid (PAA, CH₃C(O)OOH) (Phillips et al., 2012, 2013, 2016; Crowley et al., 2018).

PAN (and other peroxycarboxylic nitric anhydrides) are formed in the atmosphere via the reaction of NO₂ with peroxyacyl radicals (RC(O)O₂) generated during the photo-oxidation of volatile organic compounds. Slusher et al. (2004) reported the first detection of PAN, PPN (peroxypropionic nitric anhydride), MPAN (peroxymethacrylic nitric anhydride) and PiBN (peroxyisobutyric nitric anhydride) using thermal decomposition chemical ionization mass spectrometry (TD-CIMS) with iodide ions. The most abundant, PAN, is of great importance owing to its role in transportation of NO₂ from source regions to remote areas (Moxim et al., 1996). The detection of PAN and its analogues via I-CIMS requires thermal dissociation (generally at temperatures close to 100 ^∘C) to the peroxy radical, which then reacts with I⁻ primary ions to form the carboxylate anion, which is then detected. Compared to gas chromatographic methods for detection of peroxycarboxylic nitric anhydrides, the I-CIMS technique allows faster measurements with comparable sensitivity and selectivity (Slusher et al., 2004; Roiger et al., 2011), enabling eddy covariance flux measurements (Turnipseed et al., 2006; Wolfe et al., 2009). Although the bond dissociation energy of the peroxycarboxylic nitric anhydrides is similar, Zheng et al. (2011) report lower sensitivity for APAN (peroxyacrylic nitric anhydride), PiBN, PnBN (peroxy-n-butyric nitric anhydride) and CPAN (peroxycrotonyl nitric anhydride) and Mielke and Osthoff (2012) report lower sensitivity for MPAN compared to PPN and PAN. Peroxycarboxylic nitric anhydrides have been measured using I-CIMS in various locations including boreal forests (Phillips et al., 2013), pine forests (Turnipseed et al., 2006; Wolfe et al., 2009), urban areas (Slusher et al., 2004; LaFranchi et al., 2009; Wang et al., 2017) and the Arctic (Roiger et al., 2011).

Detection of PAA by I-CIMS was reported by Phillips et al. (2013), who performed the first combined measurement of PAN and PAA in a boreal forest in Finland. PAA acts as a significant sink for CH₃C(O)O₂ and HO₂ under low-NO_x conditions and can compete with PAN formation, especially at high temperatures (Crowley et al., 2018). As for PAN, PAA was detected as the acetate anion at a mass-to-charge ratio (m∕z) of 59. Unlike other TD-CIMS instruments that describe an absence of a residual signal when NO is added to the inlet (e.g. Warneke et al., 2016), the I-CIMS deployed by Phillips et al. (2013) and the instrument presented in this study are very sensitive to PAA at m∕z 59. Furgeson et al. (2011) also describe an interference at m∕z 59 that is not titrated by NO and suggest detection of PAA, which is produced in their photochemical source used for PAN generation. In addition, Veres et al. (2015) report a very similar mechanism to PAA detection for the detection of pernitric acid (PNA). Differences in the sensitivities of various I-CIMS instruments to PAA at m∕z 59 are likely to be associated with different de-clustering potentials. At m∕z 59, the I-CIMS deployed by Phillips et al. (2013) was insensitive to acetic acid. A wide range of organic acids can be detected as an I⁻ cluster, with the parent molecule (Le Breton et al., 2012; Lee et al., 2014) using time-of-flight mass spectrometers (I-CIMS-TOF), which have a high mass resolution and exploit the iodine mass defect for the identification of the elemental composition of the organic trace gases detected.

ClNO₂ is formed in the heterogeneous reaction of N₂O₅ on chloride-containing particles and surfaces during the night (Behnke et al., 1997). The daytime photolysis of ClNO₂ results in the release of chlorine atoms, which enhance oxidation rates of organic trace gases, especially during early morning hours (Phillips et al., 2012; Riedel et al., 2012). Nitryl chloride has been observed by I-CIMS as ${\mathrm{IClNO}}_{\mathrm{2}}^{-}$ and ICl⁻ in the polluted marine boundary layer (Osthoff et al., 2008; Riedel et al., 2012) close to the coast, for example, in Hong Kong (Wang et al., 2016) and London (Bannan et al., 2015) but also inland in continental North America (Thornton et al., 2010; Mielke et al., 2011; Faxon et al., 2015), rural continental Europe (Phillips et al., 2012) and northern China (Tham et al., 2016).

Most I-CIMS systems in operation for atmospheric measurement use a radioactive ion source (usually ²¹⁰polonium, an α-emitter) to generate the primary iodide ions from methyl iodide (CH₃I). Although this type of ion source is well-established and known for its high stability and low chemical background, important and sometimes unsurmountable obstacles to its use are the safety regulations for the shipment, storage and operation of radioactive devices containing polonium. Potential alternatives are corona discharge and X-ray ion sources, as commonly used in atmospheric pressure chemical ionization mass spectrometers (AP-CIMS) (Jost et al., 2003; Skalny et al., 2007; Kürten et al., 2011; Zheng et al., 2015), though the former have not been used for iodide ion generation.

We have developed a CI-QMS instrument (chemical ionization quadrupole mass spectrometer) with an electrical discharge ion source that generates iodide ions without the use of a radioactive ionizer. Although this instrument was originally intended for measurement of PAN, PAA and ClNO₂, we discovered that a wider variety of gas-phase species, including SO₂, HCl and acetic acid could be detected. In the following we show that the instrument is suitable for measurement (at the tens of pptv level) of trace gases connected with sulfur and chlorine chemistry, for example, in the anthropogenically influenced marine boundary layer. Its deployment as a PAN detector is limited to environments where PAN mixing ratios regularly exceed 100 pptv or when high temporal resolution is not necessary.

Figure 1 Schematic diagram of the CI-QMS. The air is sampled through the TDR (heated inlet) and enters the IMR after optional bypassing through the scrubber and mixing with calibration gas or nitrogen oxide (NO) for PAN background measurement. Ions are guided to the detector via the CDC, OCT and QMF. Typical flows (F) and pressures (p) are F₁=2.2, F₂=1.0, F₃=0.8 slm, p₁= ambient pressure, p₂=24, p₃=0.6, $p_{\mathrm{4}}=\mathrm{6}\times {\mathrm{10}}^{-\mathrm{3}}$, and $p_{\mathrm{5}}=\mathrm{9}\times {\mathrm{10}}^{-\mathrm{5}}$ mbar.

2 Instrumentation

Our chemical ionization quadrupole mass spectrometer (CI-QMS) is based on the thermal dissociation technique described by Slusher et al. (2004) and Zheng et al. (2011) and was originally constructed in collaboration with Georgia Tech as a prototype THS Instruments product. A schematic diagram of the instrument in its present form is given in Fig. 1. The major modification, forced by issues of the restricted use of polonium on some platforms, is the replacement of the ²¹⁰Po ionizer with an electrical discharge ion source (see Fig. 2). The instrument as sketched in Fig. 1 consists of a thermal dissociation region (TDR), discharge ion source (DIS), ion molecule reactor (IMR), collisional dissociation chamber (CDC), octopole ion guide (OCT), quadrupole mass filter (QMF) and detector (DET). The four different vacuum chambers are separated by critical orifices and pumped by a combination of scroll pumps and turbo-molecular pumps. The CI-QMS is built into an aircraft rack (HALO, Gulfstream G550) to facilitate airborne operation. A description of the various parts of the instrument as identified in Fig. 1 is given in the following.

Figure 2 Schematic drawing of the RF discharge ion source. The high voltage and the distance between the tips of the tungsten electrodes are variable, as described in the text. A transformer, which is supplied with up to 200 V AC from the internal V25 unit, applies temporarily fluctuating potentials to both electrodes. The body of the ion source (coloured in light brown) is made of stainless steel. PVDF is polyvinylidene fluoride. The region around the electrodes can optionally be pumped as described in the text.

2.1 Thermal dissociation region (TDR)

During standard operation, a flow (F₁) of 2.2 L (standard pressure and diameter, STD) min⁻¹ (slm) is sampled through the instrument inlet (see Fig. 1). Ambient air entering the TDR first passes through a 2 µm pore size membrane filter (Pall Teflon) to efficiently remove particles. The thermal decomposition of PAN takes place in a 20 cm length of perfluoroalkoxy (PFA) tubing (9.3 mm internal diameter, ID) enclosed in a snugly fitting stainless steel shell heated to 200 ^∘C. A gas temperature of 160–170 ^∘C was measured inside the heated section of tubing. The last 4 cm long section of PFA in front of the orifice to the IMR is not actively heated but the gas temperature still remains at ≈130 ^∘C, suppressing PAN recombination. Due to space restrictions inside the aircraft rack, the oven is curved over a 90^∘ bend; however, no significant reduction in sensitivity due to a loss of CH₃C(O)O₂ on the PFA walls could be observed when compared to a straight TD inlet. A bypass flow of 1 slm (F₂) decreases the inlet residence time and therefore optimizes the peroxyacyl radical transmission. However, as PAN is calibrated in situ with a photochemical source (see Sect. 2.7), the fractional transmission of CH₃C(O)O₂ does not need to be known. A total of 1.2 slm (F₁ minus F₂) of the inlet flow enters the IMR via a constant-pressure orifice (see Sect. 2.3) and mixes with the 0.8 slm flow (F₃) of CH₃I∕N₂ passing through the ion source. For ground-level deployment, the TDR is held at ambient pressure resulting in a residence time of ≈200 ms at 1 bar. For application in an aircraft, the arrangement of the TDR and constant-pressure orifice can be switched so that a constant pressure of p₁=100 hPa is established in the TDR, resulting in a 40 ms residence time. With this mode of operation, measurements can be made at altitudes up to ≈15 km depending on aircraft inlet configuration.

2.2 Discharge ion source (DIS)

The radio-frequency (RF) discharge ion source represents the major difference to I-CIMS instruments commonly described in the literature. It consists of two tungsten needles, their tips placed at a distance of approximately 6 mm (adjustable) from each other (see Fig. 2); changing this distance by a few millimetres did not have a large effect on the overall ion count rate. The vacuum fittings through which the needles enter the discharge volume can optionally be evacuated to eliminate an air leak into the ion source. A 2.5 kV voltage (20 kHz) applied across the tungsten needles leads to the formation of a stable glow discharge (3 mA). An RF discharge was chosen for its advantages for operation in negative ion mode and because it is easier to handle with regard to electric field geometry and polarity. The discharge between the tungsten needles can be observed by eye or spectroscopically through a quartz viewing port; a photograph of the glow and the dispersed N₂ emission spectrum due to the B³Π_g← C³Π_u transition (Lofthus and Krupenie, 1977; Bayram and Freamat, 2012) is shown in Fig. S1 of the Supplement. The line intensities increase with the voltage applied (≈ factor of 2.5 from 1500 to 3000 V), but the relative intensities do not change significantly. From the relative line intensities we calculate that the N₂ molecules have a vibrational temperature of ≈3000 K (Svarnas, 2013).

In normal operation, a flow of 0.8 slm (F₃) of 2 ppmv methyl iodide (CH₃I) in nitrogen passes through the ionization region. In order to prevent the backflow of air from the IMR into the discharge region, the flow through the ionization region is kept high (0.8 slm) and passes through a 0.9 mm aperture before entering the IMR.

Similar to the α radiation of a ²¹⁰Po ionizer, I⁻ ions are formed via dissociative electron attachment to CH₃I. In contrast to typical corona discharge ion sources with a current of a few µA (see, e.g. Kürten et al., 2011), our glow discharge operates at ≈3 mA, which leads to highly energetic electrons and ions in the ionization region and a more complex mass spectrum (see Sect. 3).

2.3 Ion molecule reactor (IMR)

Under standard operating conditions for ground-level measurements, a flow of 1.2 slm of the air to be analysed (F₁ minus F₂) is mixed with the 0.8 slm flow of CH₃I∕N₂ (F₃) entering the IMR from the ion source. The IMR is evacuated by a dry scroll vacuum pump (ULVAC DISL-101, 100 L min⁻¹) and is held at a constant pressure of 24 mbar (p₂). For operation above the boundary layer, an extra 50 cm³ (standard temperature and pressure, STP) min⁻¹ (sccm) flow of humidified air is added just in front of the IMR to ensure that sufficient water vapour is present to form I⁻(H₂O) clusters. The role of I⁻(H₂O) and other primary-ion clusters with water is discussed in Sect. 5.

The constant pressure within the IMR is achieved by use of a variable orifice consisting of two metal plates, one with a hole shaped like the 2-D projection of a bike saddle and one with a circular hole, the relative position of which (i.e. the degree of overlap of the holes) is controlled by a stepper motor. The saddle form was chosen as it results in a roughly linear relationship between stepper motor position and mass flow through the orifice, enabling rapid and precise adaptation to changes in ambient pressure even during steep ascents or dives of an aircraft.

The temperature in the IMR is above ambient owing to the inflow of heated gas through the TDR. The exact residence time for trace gases to react with primary ions in the IMR is not accurately known, as the mixing of the gas flows and temperature evolution in the IMR are not well characterized. Based on the mass flow rate into the IMR and its volume (≈100 cm³) and disregarding ion drift due to the potential applied between the IMR and the CDC, we calculate an approximate reaction time of ≈70 ms.

2.4 Collisional dissociation chamber (CDC) and octopole ion guide (OCT)

The CDC region consists of an octopole ion guide to accelerate and collimate the effusive ion beam entering from the IMR. It is separated from the IMR by a critical orifice (0.8 mm diameter) and held at a pressure of 0.6 mbar (p₃) by the Holweck stage of a turbo-molecular pump (Leybold Turbovac 90 i, 90 L s⁻¹ with an Agilent IDP-3 scroll pump, 50 L min⁻¹ as a backup pump). A potential difference of typically 20 V is applied in the CDC, which results in the de-clustering of weakly bound adducts (often with H₂O), a simplified mass spectrum and a higher sensitivity to the product ion of interest. The de-clustering voltage can be varied independently for each ion of interest, thus optimizing sensitivity for individual trace gases. An example of the variable de-clustering potential during operation of the CI-QMS in the selected ion monitoring mode (e.g. to measure the I⁻(H₂O) cluster or differentiate between acetic and peracetic acid) is given below.

An additional octopole ion guide in the subsequent vacuum chamber ($\mathrm{6.0}\times {\mathrm{10}}^{-\mathrm{3}}$ mbar, p₄) further collimates the ion beam and guides it to the detector region. This octopole ion guide is evacuated by the turbo-molecular stage of the Leybold Turbovac 90 i.

2.5 Quadrupole mass filter (QMF) and detector (DET)

A radio-frequency generator (Balzers QMH 410-3, 1.44 MHz) provides a combination of direct and alternating voltage to the quadrupole rods (10 mm) so that ions with a specific m∕z are forced on stable trajectories and reach the detector. These ions are then detected with a channel electron multiplier (ITT Ceramax 7550M). The detector chamber is pumped to a pressure of $\mathrm{9.0}\times {\mathrm{10}}^{-\mathrm{5}}$ mbar (p₅) by a turbo-molecular pump (Varian V70LP, 70 L s⁻¹).

2.6 Scrubber

To determine the instrumental and chemical background, the sampled air is automatically and periodically bypassed into a scrubber (see Fig. 1) consisting of a 20 cm long stainless steel oven filled with steel wool and heated to 120 ^∘C. The trace gases discussed in this work are all destroyed efficiently by the hot metal surfaces whilst leaving the relative humidity unaffected.

2.7 Photochemical PAN source

For in situ PAN calibration we use a photochemical source based on the method of Warneck and Zerbach (1992) but with a phosphor-coated Pen-Ray mercury lamp (Jelight, broad emission centred at 285 nm) as described by Flocke et al. (2005). Typically, 50 sccm of acetone (200 ppmv in synthetic air, Air Liquide) and 5 sccm of NO (1 ppmv in N₂, Air Liquide) are mixed in a quartz glass reactor (150 mL volume, actively cooled by a fan) at 1050 mbar forming PAN (Reactions R1–R4). The calibration source converts NO almost stoichiometrically to PAN (conversion > 95 %) and results in a mixing ratio of about 4 ppbv of PAN in the TDR. The PAN source is continuously operated and its 55 sccm output drains into the exhaust line, periodically switching into the main flow during scrubbing. The conversion efficiency of NO to PAN was checked using thermal dissociation cavity ring-down spectroscopy as described previously (Phillips et al., 2013). The PAN source also generates both PAA and acetic acid (Reactions R5–R6), which, as described below, are also detected by the CI-QMS.

$$\begin{array}{}\text{(R1)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{CH}}_{\mathrm{3}}+h\mathit{\nu }\left({\mathrm{O}}_{\mathrm{2}}\right)\to {\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}+{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}\text{(R2)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}+\mathrm{NO}\left({\mathrm{O}}_{\mathrm{2}}\right)\to \mathrm{HCHO}+{\mathrm{HO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}\text{(R3)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+\mathrm{NO}\left({\mathrm{O}}_{\mathrm{2}}\right)\to {\mathrm{CH}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{2}}+{\mathrm{CO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}\text{(R4)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}+\mathrm{M}\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}{\mathrm{NO}}_{\mathrm{2}}+\mathrm{M}\text{(R5)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right)\mathrm{OOH}+{\mathrm{O}}_{\mathrm{2}}\text{(R6)}& {\displaystyle }& {\displaystyle }\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right)\mathrm{OH}+{\mathrm{O}}_{\mathrm{3}}\end{array}$$

2.8 Electronics and data acquisition

The vast majority of the instrument's electronics are controlled by a “V25” system developed in-house. The V25 handles the interplay between single components such as flow controllers, pressure gauges, magnetic valves, thermocouples, heaters, MS potentials and RF generators. All command sequences and measurement cycles (background, calibration, etc.) can be customized and fully automated for operation in aircraft or in remote locations. During measurement campaigns we usually focus on specific trace gases and operate in selected ion monitoring mode, typically monitoring between 3 and 10 different values of m∕z. Different m∕z values can be adjusted in a few milliseconds by variation in the direct voltage and the amplitude of the alternating voltage applied to the quadrupole rods. The integration time of the detector for a single channel is usually set to 10 ms for primary ions and 100 ms for product ions, which represents a compromise between high signal-to-noise ratio (S∕N) and high temporal resolution. For each m∕z monitored, the integrated signal is calculated by summing up eight individual channels, resulting in a total integration time (for one product ion) of about 800 ms. Higher-frequency measurements are possible at the cost of a reduction in the signal-to-noise ratio. The counts for each channel and the integrated counts as well as the most important system parameters are saved on an internal PC card and can additionally be collected and monitored online using customized LabView software. To identify additional traces gases of interest, the whole mass spectrum (m∕z 1–256) is occasionally scanned and recorded, which takes about 1–2 min.

2.9 Size, weight and power consumption

The CI-QMS is situated in a compact aircraft rack (65 cm×55 cm×140 cm, HALO, Gulfstream G550) with a total weight of 135 kg and a power consumption of 0.9 kW with the vacuum pumps as the main power consumers. The two vacuum scroll pumps require 230 V AC input whereas all the other components are operated with 24 V DC from an AC/DC converter that either can be supplied with 230 V AC or 115 V AC (three phases, 400 Hz, for aircraft operation).

Table 1 Ion source configurations according to Fig. 3 and primary ions observed.

3 Primary-ion spectra

The deployment of an RF discharge source for iodide ion production leads to a more complex primary-ion mass spectrum when compared to use of ²¹⁰Po; consequently, a wider variety of trace gases can be detected. Here we compare both ion sources with respect to sensitivity and achievable detection limits of a number of trace gases. Figure 3 illustrates the primary-ion spectrum with a discharge ion source under various conditions and compares it to that obtained using ²¹⁰Po. The absolute ion count rates for both ion sources are comparable, i.e. up to ($\mathrm{6}\pm \mathrm{2})\times {\mathrm{10}}^{\mathrm{6}}$ Hz for I⁻. Details of the configurations (i–v) and primary ions observed are summarized in Table 1.

Figure 3 Primary-ion spectra (using I₂∕N₂ (v) or CH₃I∕N₂ (iii–iv) and using either pure synthetic air (iii) or N₂ (iv) for the inlet gas) obtained using the RF discharge (iii–v) or ²¹⁰Po (i–ii) as an ion source. The integrated counts (eight channels) for each m∕z normalized to the highest peak present are shown, which is I⁻. For this reason, true peak shapes are not visible. A description of the different configurations can be found in Table 1.

The primary-ion mass spectrum obtained by passing CH₃I∕N₂ through the ²¹⁰Po ionizer (370 MBq, configuration i) at typical relative humidity (50 % at 25 ^∘C) and low de-clustering voltage (0–2 V) in the CDC is dominated by I⁻ and I⁻(H₂O) at m∕z 127 and 145, with no other significant ion peaks present at > 0.1 % relative signal strength to I⁻. The background signal for all trace gases of interest, i.e. PAN and PAA at m∕z 59 and ClNO₂ at m∕z 208 and 210, is consequently negligible and the detection limits are correspondingly low (a few pptv in 1 s integration time).

With RF discharge and CH₃I∕N₂ as an ion source gas (configuration iii, applied in all the field measurements we discuss below), the primary-ion mass spectrum is more complex with additional ions such as CNO⁻ (m∕z 42, 37 % of I⁻ at low de-clustering), ${\mathrm{CO}}_{\mathrm{3}}^{-}$ (m∕z 60, 32 % of I⁻), ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (m∕z 62, 5 % of I⁻), ${\mathrm{IO}}_{\mathrm{3}}^{-}$ (m∕z 175, 28 % of I⁻) and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ (m∕z 179, 67 % of I⁻). With the de-clustering voltage set to 20 V (the best S∕N ratio for most molecules of interest), the background count rate compared with ²¹⁰Po is elevated by at least 1 order of magnitude for m∕z 59, which is used to monitor PAN and PAA. The high chemical background is assumed to originate from CH₃I breakdown in the discharge ion source and formation of ${\mathrm{O}}_{\mathrm{2}}^{-}$ in the ion source and IMR. When pumping the region around the ion source needles, the formation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (m∕z 62) can be reduced by a factor of 2 but the other ions still show similar ion count rates. This observation suggests that a small amount of O₂ entering the ion source can form additional ${\mathrm{NO}}_{\mathrm{3}}^{-}$. According to manufacturer's specifications the nitrogen supply (N₂ 6.0, Westfalen AG) can contain up to 0.5 ppmv O₂ and H₂O, which can also result in the formation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in the ion source.

Figure 4 Ion detection schemes for SO₂, PAN, PAA, acetic acid, ClNO₂ and HCl using the RF discharge ion source.

To examine the influence of oxygen in the IMR on the primary ions formed, we switched the main gas flow (i.e. that which does not pass through the RF discharge) from ambient air to pure nitrogen (configuration iv). In this case, apart from I⁻, only CN⁻ (m∕z 26, 2 % of I⁻ at low de-clustering), IH(CN)⁻ (m∕z 154, 6 % of I⁻) and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ (m∕z 179, 9 % of I⁻) remained. While these conditions are unrealistic for atmospheric measurements they clearly indicate that the presence of additional primary ions containing O atoms and the elevated chemical background on m∕z of interest are highly dependent on the amount of O₂ present in the IMR. With pure N₂ in the inlet (configuration iv), background signals at m∕z 59, 188, 207 and 208 could be lowered by about an order of magnitude. The use of N₂ results in a drastically reduced sensitivity to SO₂, as the primary ion used to detect it (${\mathrm{IO}}_{\mathrm{3}}^{-}$; see Sect. 4.2) is no longer abundant. This is illustrated in Fig. S2, where we plot the dependence of the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ signal (m∕z 175) on the fractional pressure of O₂ in the inlet and the signal at m∕z 207 (ISO${}_{\mathrm{3}}^{-}$) used to monitor SO₂ (see Sect. 4.2). Clearly, detection of SO₂ is not possible without the presence of O₂ and is not available when using ²¹⁰Po as an ion source.

In an attempt to improve the detection limit for PAN by lowering the background signal on m∕z 59, I₂, produced from a flow of nitrogen over iodine crystals (configuration v), was used instead of CH₃I. This resulted in a significant reduction of background signals, especially on m∕z 59 and the disappearance of all the ion peaks containing C and N atoms with just ${\mathrm{IO}}_{\mathrm{3}}^{-}$ (m∕z 175, 51 % of I⁻), ${\mathrm{IO}}_{\mathrm{4}}^{-}$ (m∕z 191, 38 % of I⁻), ${\mathrm{IO}}_{\mathrm{2}}^{-}$ (m∕z 159, 12 % of I⁻) and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (m∕z 62, 8 % of I⁻) remaining. However, use of I₂ was accompanied by a drastic lowering of the sensitivity to PAN, despite comparable I⁻ ion counts at m∕z 127. The decrease in sensitivity can be traced back to equilibrium between I⁻, I₂ and ${\mathrm{I}}_{\mathrm{3}}^{-}$.

$$\begin{array}{}\text{(R7)}& {\displaystyle }{\mathrm{I}}^{-}+{\mathrm{I}}_{\mathrm{2}}+\mathrm{M}\leftarrow \to {\mathrm{I}}_{\mathrm{3}}^{-}+\mathrm{M}\end{array}$$

An equilibrium constant (K_eq, in bar⁻¹) of K_eq= [${\mathrm{I}}_{\mathrm{3}}^{-}$]/[I⁻][I₂]= exp(11300/T) for Reaction (R7) (based on the Gibbs free energy of −94.14 kJ mol⁻¹; NIST Webbook, 2010) and an estimate (based on its saturation vapour pressure) of the concentration of I₂ of $\approx \mathrm{2}\times {\mathrm{10}}^{-\mathrm{7}}$ bar results in the complete dominance (by several orders of magnitude) of [${\mathrm{I}}_{\mathrm{3}}^{-}$] compared to [I⁻] in the IMR. While the presence of large concentrations of ${\mathrm{I}}_{\mathrm{3}}^{-}$ may explain the large signal at m∕z 127 following de-clustering, we conclude that the reaction between ${\mathrm{I}}_{\mathrm{3}}^{-}$ and CH₃C(O)O₂ is very inefficient or does not lead to ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ formation. I₂ does not represent a feasible alternative to CH₃I for PAN measurement and HCl detection is not possible. However, we can still detect SO₂ via ${\mathrm{IO}}_x^{-}$ primary ions and also acetic acid, presumably due to (${\mathrm{IO}}_x^{-}$) clusters with CH₃C(O)OH.

In another experiment performed with the ²¹⁰Po source, synthetic air instead of nitrogen was flowing over the polonium ionizer (configuration ii), simulating a huge leak of oxygen into the source. Besides I⁻ and I⁻(H₂O), ${\mathrm{O}}_{\mathrm{2}}^{-}$ (m∕z 32, 5 % of I⁻), ${\mathrm{O}}_{\mathrm{2}}^{-}\left({\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right)$ (m∕z 50, 3 % of I⁻) and ${\mathrm{CO}}_{\mathrm{3}}^{-}$ (m∕z 60, 2 % of I⁻) were present, but ions like ${\mathrm{IO}}_x^{-}$ and $\mathrm{I}(\mathrm{CN}{)}_x^{-}$ that are probably responsible for the detection of SO₂ and HCl (see below) were not observed as they are unique to the RF discharge ion source. In combination with the experiment where the housing around the tungsten needles was evacuated, we conclude that a leak of O₂ into the discharge region might increase ${\mathrm{O}}_{\mathrm{2}}^{-}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ but is not responsible for the complex primary-ion spectrum observed with our discharge ion source. No change in the ion spectrum was observed when the linear steel tubing between the ion source and IMR was replaced by tubing with a 90^∘ bend. This result precludes an important role for ion formation via highly energetic radiation from the ionization region reaching the IMR interacting with O₂. The diffusion of oxygen from the IMR into the ion source is also very unlikely due to a high-volume flow between the ion source and IMR and the use of a small aperture. We conclude that the role of O₂ in the formation of primary ions containing I, O, C and N atoms in the IMR is most likely related to its role as a trapper and carrier of electrons, possibly as excited ${\mathrm{O}}_{\mathrm{2}}^{-}$ anions.

4 Detection schemes and calibration methods

Figure 4 shows which trace gases we can detect with our instrument using the RF discharge ion source. The middle branch (b) represents ionization via I⁻ and its water cluster ion and is the same as ²¹⁰Po-based ion generation schemes frequently used for detection of PAN (Slusher et al., 2004), PAA (Phillips et al., 2013) and ClNO₂ (McNeill et al., 2006). The outer branches (a and c) are unique to our CI-QMS using the discharge ion source and can be attributed to the presence of different primary ions. The presence of H, C and N atoms from CH₃I∕N₂ breakdown in the discharge region leads to the existence of the right-hand branch (c) that disappears when CH₃I is replaced by I₂ (see Sect. 3, configuration v). The usual presence of oxygen inside the IMR is responsible for the existence of the left-hand branch (a) and is not available with configuration (iv) in which nitrogen was used for the main gas flow instead of synthetic air. For the combined detection of PAN, ClNO₂, SO₂, HCl, peracetic and acetic acid, configuration (iii) was used in all of our field measurements. In the following we discuss in detail the ion–molecule reactions involved in the detection of these trace gases and also outline how the CI-QMS is calibrated and which sensitivities and detection limits can be achieved.

Table 2 Sensitivity and limit of detection of the CI-QMS for PAN, PAA, acetic acid, ClNO₂, SO₂ and HCl.

^a Sensitivity S (in Hz ppt⁻¹) at 50 % RH (25 ^∘C), normalized to 10⁶ Hz I⁻ with de-clustering set to 20 V. ^b Limit of detection (LOD, 2σ, 1 s integration time). ^c The LOD is calculated from the background signal at m∕z 59 and does not include the reduction in detection limit incurred when acetic acid and PAA are present at the same m∕z as described in the text. ^d Without de-clustering applied.

4.1 PAN, PAA and acetic acid

PAN, PAA and acetic acid are all detected as ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ at m∕z 59. Depending on the relative humidity and the de-clustering potential in the CDC, clusters of ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ with H₂O are observed at m∕z 77. Sensitivities and product yields for the detection of these three molecules are summarized in Table 2. The detection mechanism for PAN using I⁻ primary ions is the same as that reported when using ²¹⁰Po as an ion source (Slusher et al., 2004). PAN is thermally decomposed inside the TDR into a peroxy radical (CH₃C(O)O₂) and NO₂ via Reaction (R8). The rate coefficient for the thermal decomposition of PAN (at 453 K and 1 bar) is ≈2000 s⁻¹ (Atkinson et al., 2006; IUPAC, 2018), so that > 99.99 % of PAN should be thermally dissociated within 200 ms. This could be confirmed by measurement of the signal due to a stable PAN source whilst varying the inlet temperature. The CH₃C(O)O₂ product reacts with I⁻ in the IMR to form ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ (m∕z 59) via Reaction (R9) involving clusters with water vapour. The detection of PAA with I⁻ primary ions is believed to be direct via Reaction (R10) (Phillips et al., 2013).

$$\begin{array}{}\text{(R8)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}{\mathrm{NO}}_{\mathrm{2}}+\mathrm{M}\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}+\mathrm{M}\text{(R9)}& {\displaystyle }& {\displaystyle }{{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+{\mathrm{I}}^{-}\left({\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right)}_n\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}^{-}{\left({\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right)}_n+\mathrm{IO}\text{(R10)}& {\displaystyle }& {\displaystyle }{{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right)\mathrm{OOH}+{\mathrm{I}}^{-}\left({\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right)}_n\to {\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}^{-}{\left({\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right)}_n+\mathrm{HOI}\end{array}$$

When using the RF discharge ion source there is also an additional pathway for PAN and PAA detection, involving $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ primary ions, resulting in the formation of I(CN)${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$, which is observed at m∕z 212 when de-clustering is switched off. With de-clustering this ion fragments to m∕z 59. However, the sensitivity is relatively low, and the selectivity is not improved as acetic acid is also detected at this m∕z (see Table 2).

The separation of PAN from PAA and acetic acid signals when sampling air masses which contain both trace gases can be achieved by cooling the TDR to prevent formation of CH₃C(O)O₂ and thus detection of PAN (Phillips et al., 2013) or by adding NO to the TDR in order to remove CH₃C(O)O₂ (Reaction R11).

$$\begin{array}{}\text{(R11)}& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right){\mathrm{O}}_{\mathrm{2}}+\mathrm{NO}\to {\mathrm{CH}}_{\mathrm{3}}+{\mathrm{CO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}\end{array}$$

The latter method has the advantage of being more rapid, as NO can be switched in and out of the TDR in a matter of seconds, whereas cooling of the inlet may take minutes. Generally we add NO (100 ppmv in nitrogen, Air Liquide) to the TDR at a mixing ratio of 0.23 ppmv (3.7×10¹² molecule cm⁻³). Given the approximate residence time of ca. 200 ms in the TDR (calculated from the volume of the TDR and the volumetric flow rate) and the rate coefficient for Reaction (R11) of $\mathrm{1.4}\times {\mathrm{10}}^{-\mathrm{11}}$ cm³ molecule⁻¹ s⁻¹ at 453 K (Atkinson et al., 2004; IUPAC, 2018), we calculate that more than 99.99 % of the CH₃C(O)O₂ is removed by titration with NO. When NO is added, we therefore measure only PAA (or the sum of PAA and acetic acid, depending on de-clustering potential; see below). The signal due to PAN is then calculated by subtracting the interpolated signal during NO addition.

The use of the RF discharge source also results in sensitivity to acetic acid at m∕z 59, which is not observed using ²¹⁰Po as an ion source. Rather, detection of acetic acid has been reported at m∕z 187 (ICH₃C(O)OH⁻) (Lee et al., 2014). Our instrument is relatively insensitive for acetic acid at m∕z 187 and, without de-clustering, we measure acetic acid mainly as (CNO)CH₃C(O)OH⁻ (m∕z 102) and I(CN)${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ (m∕z 212).

When monitoring m∕z 59 using the discharge ion source, the background signal during NO addition consists of both PAA and acetic acid. To differentiate between them we make use of the fact that the relative sensitivity to PAA and acetic acid at m∕z 59 depends on the de-clustering potential applied in the CDC. We find that acetic acid is only detected at m∕z 59 when a de-clustering voltage of about 20 V is applied, whereas PAA is detected at m∕z 59 both with and without de-clustering, albeit with different sensitivity. The difference is related to the fact that for PAA the product ion is formed directly in Reaction (R10), whereas acetic acid is believed to initially form a cluster with $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ (Reaction R12) and only dissociate to ${\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}$ when the 20 V de-clustering voltage is applied (Reaction R13). ICN⁻, IH(CN)⁻ and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ are all potential primary ions for detection of acetic acid, though $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ is the most abundant. When NO is added, the mixing ratio of acetic acid can be calculated by subtracting the signal without de-clustering from the signal with de-clustering. Unfortunately, the chemical background without de-clustering at m∕z 59 is a factor of 2.5 higher than with de-clustering and the sensitivity is reduced, which increases the LOD for detection of PAA significantly (see Table 2).

$$\begin{array}{}\text{(R12)}& {\displaystyle }& {\displaystyle }{\mathrm{CH}}_{\mathrm{3}}\mathrm{C}\left(\mathrm{O}\right)\mathrm{OH}+\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}\to \mathrm{HCN}+\mathrm{I}\left(\mathrm{CN}\right){\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}\text{(R13)}& {\displaystyle }& {\displaystyle }\mathrm{I}\left(\mathrm{CN}\right){\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}+\mathrm{\Delta }U\to \mathrm{ICN}+{\mathrm{CH}}_{\mathrm{3}}{\mathrm{CO}}_{\mathrm{2}}^{-}\end{array}$$

Results of combined PAA and acetic acid measurements from the CYPHEX field campaign and speciation via changing the CDC parameters can be found in Derstroff et al. (2017). An exemplary time series showing the different measurement modes of the CI-QMS (scrubber, ambient, NO addition) and the differences in detection of PAA and acetic acid when applying a de-clustering voltage is provided in Fig. S3.

An enhancement in sensitivity to acetic acid at m∕z 59 when using the RF discharge ion source compared to ²¹⁰Po is illustrated by the signals obtained from the photochemical source used to generate PAN, which also generates unquantified amounts of both PAA and acetic acid (Sect. 2.7, Reactions R5 and R6). When using the RF discharge ion source and 20 V de-clustering voltage, the signal ratio of PAN ∕ (PAN + PAA + acetic acid) at m∕z 59 is ≈0.2. In contrast, using ²¹⁰Po as an ion source the ratio is ca. 0.9. As the relative sensitivity (m∕z 59) to PAN and PAA is similar, this change in ratio reflects enhanced instrument sensitivity to acetic acid when using the discharge ion source with de-clustering. This represents a significant disadvantage of the RF discharge source for PAN detection compared to ²¹⁰Po. Not only is the instrumental background at m∕z 59 higher, the presence of acetic acid in ambient air samples means that a larger and more variable chemical background signal has to be subtracted to calculate the PAN mixing ratio, which increases the limit of detection and overall uncertainty significantly.

Calibration of PAN, PAA and acetic acid

The in situ calibration of PAN is described in Sect. 2.7. The overall uncertainty of the calibration, based on the uncertainty in dilution, the mixing ratio of the NO calibration cylinder (1 ppmv) and the conversion efficiency from NO to PAN is ≈10 %.

For PAA, two methods, both using a diffusion source containing a commercially available 39 % solution of PAA in acetic acid, have been used to calibrate the CI-QMS. In the first, we use simultaneous CI-QMS and wet-chemical peroxide-specific detection of PAA based on the horseradish peroxidase/catalase/p-hydroxyphenyl fluorescence measurement technique (Lazrus et al., 1986) in which organic peroxides (and peracids) are converted to H₂O₂ (AL2021, Aero-Laser GmbH). The wet chemical method is calibrated via standard H₂O₂ solutions and the overall uncertainty (related to scrubbing efficiency of PAA) is 13 %. As the AL2021 is not always available during the campaign, we developed a second approach in which PAA undergoes transformation to ${\mathrm{I}}_{\mathrm{3}}^{-}$ (aq), which can be quantified using aqueous-phase absorption spectroscopy (Awtrey and Connick, 1951; Friedrich, 2015). Based on uncertainty in the scavenging of PAA into an acidified, aqueous solution, uncertainty associated with the absorption cross section of ${\mathrm{I}}_{\mathrm{3}}^{-}$ and the reproducibility of ${\mathrm{I}}_{\mathrm{3}}^{-}$ signals when sampling from a constant source of PAA, we estimate the total uncertainty of the ${\mathrm{I}}_{\mathrm{3}}^{-}$ method to be ≈30 %.

For the calibration of acetic acid we use a permeation source (7.33 ng min⁻¹ at 30 ^∘C, Metronics) with an uncertainty of 8 %. Additionally, we sampled the output of a diffusion source of pure liquid acetic acid simultaneously using the CI-QMS (41.2 ppbv after dilution) and an infrared absorption spectrometer measuring CO₂ (LI-COR) following the stoichiometric thermal oxidation of acetic acid to CO₂ (Veres et al., 2010). The uncertainty of this calibration method is ≈10 %. Within combined uncertainty both methods indicated the same sensitivity of the CI-QMS to acetic acid.

4.2 Sulfur dioxide

The “standard” use of ²¹⁰Po ionization does not allow for the sensitive detection of SO₂ using I⁻ ions. As outlined in Sect. 3, additional primary ions (${\mathrm{IO}}_x^{-}$) formed with our RF discharge ion source enable SO₂ detection as, for example, ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ (m∕z 207). In addition, ${\mathrm{ISO}}_{\mathrm{4}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{-}$ and ${\mathrm{SO}}_{\mathrm{5}}^{-}$ are also formed and the relative yields are listed in Table 2. Although the underlying ion–molecule reactions resulting in their formation are not fully characterized, based on the observation of ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and ${\mathrm{IO}}_{\mathrm{4}}^{-}$ in the primary-ion mass spectrum we propose the following scheme (Reactions R14–17).

$$\begin{array}{}\text{(R14)}& {\displaystyle }& {\displaystyle }{\mathrm{SO}}_{\mathrm{2}}+{\mathrm{IO}}_{\mathrm{3}}^{-}\to {\mathrm{O}}_{\mathrm{2}}+{\mathrm{ISO}}_{\mathrm{3}}^{-}(m/z\mathrm{207})\text{(R15)}& {\displaystyle }& {\displaystyle }{\mathrm{SO}}_{\mathrm{2}}+{\mathrm{IO}}_{\mathrm{4}}^{-}\to {\mathrm{O}}_{\mathrm{2}}+{\mathrm{ISO}}_{\mathrm{4}}^{-}(m/z\mathrm{223})\text{(R16)}& {\displaystyle }& {\displaystyle }{\mathrm{SO}}_{\mathrm{2}}+{\mathrm{IO}}_{\mathrm{3}}^{-}\to \mathrm{IO}+{\mathrm{SO}}_{\mathrm{4}}^{-}(m/z\mathrm{96})\text{(R17)}& {\displaystyle }& {\displaystyle }{\mathrm{SO}}_{\mathrm{2}}+{\mathrm{IO}}_{\mathrm{4}}^{-}\to \mathrm{IO}+{\mathrm{SO}}_{\mathrm{5}}^{-}(m/z\mathrm{112})\end{array}$$

As written, Reactions (R14) and (R16) involving the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ anion are exothermic with reaction enthalpies of $\approx -\mathrm{250}$ and −113 kJ mol⁻¹, respectively. The enthalpies of formation used to derive these values were taken from the literature: ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(SO₂)  $=-\mathrm{287}$ kJ mol⁻¹ (Chase, 1998), ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(IO) =126 kJ mol⁻¹ (Goos et al., 2005) and ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(${\mathrm{SO}}_{\mathrm{4}}^{-}$) $=-\mathrm{738}$ kJ mol⁻¹ (NIST Webbook, 2010) or calculated from other thermodynamic properties. The formation enthalpy for ${\mathrm{IO}}_{\mathrm{3}}^{-}$ (ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(${\mathrm{IO}}_{\mathrm{3}}^{-}$) $=-\mathrm{211}$ kJ mol⁻¹) was calculated from its electron affinity (453.5 kJ mol⁻¹; Wen et al., 2011) and the formation enthalpy of IO₃ (242 kJ mol⁻¹; Goos et al., 2005). The formation enthalpy for ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ (ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(${\mathrm{ISO}}_{\mathrm{3}}^{-}$) $=-\mathrm{752}$ kJ mol⁻¹) was calculated from the ${\mathrm{SO}}_{\mathrm{3}}-{\mathrm{I}}^{-}$ bond strength (161 kJ mol⁻¹; Hao et al., 2005) and the formation enthalpies of I⁻ (−195 kJmol⁻¹; Goos et al., 2005) and SO₃ (−396 kJ mol⁻¹; Goos et al., 2005). In the absence of thermodynamic data for ${\mathrm{IO}}_{\mathrm{4}}^{-}$, we cannot assess the reaction enthalpies for Reactions (R15) and (R17). The iodine-containing ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ (m∕z 207) is most specific and suitable for monitoring SO₂ with good sensitivity.

Calibration of SO₂

SO₂ is calibrated by the addition of a small flow of SO₂ from a gas cylinder (1 ppmv in synthetic air, Air Liquide). The true mixing ratio of SO₂ flowing from the bottle into the absorption cell at 1 bar pressure was determined via UV-absorption spectroscopy using a white cell and diode array set-up (Wollenhaupt et al., 2000) and an absorption spectrum (290–320 nm) from the literature (Bogumil et al., 2003). The mixing ratio that was determined in this manner was found to agree to within 10 % of the manufacturers specifications. The linearity of the CI-QMS signal with the SO₂ mixing ratio (up to 60 ppbv) is shown as Fig. S4a.

4.3 Nitryl chloride

The scheme for detection of ClNO₂ using I⁻ ions generated using ²¹⁰Po is well-established (Osthoff et al., 2008; Thornton et al., 2010). Both ICl⁻ (m∕z 162 and 164) and ${\mathrm{IClNO}}_{\mathrm{2}}^{-}$ (m∕z 208 and 210) are formed, the latter generally preferred to monitor ClNO₂ in ambient air owing to potential interference through reactions of other chlorine-containing trace gases forming ICl⁻.

$$\begin{array}{}\text{(R18)}& {\displaystyle }& {\displaystyle }{\mathrm{ClNO}}_{\mathrm{2}}+{\mathrm{I}}^{-}\to {\mathrm{IClNO}}_{\mathrm{2}}^{-}\text{(R19)}& {\displaystyle }& {\displaystyle }{\mathrm{ClNO}}_{\mathrm{2}}+{\mathrm{I}}^{-}\to {\mathrm{ICl}}^{-}+{\mathrm{NO}}_{\mathrm{2}}\end{array}$$

The same ions are observed during operation with the RF discharge ion source with similar product yields for ²¹⁰Po as the ones for the RF discharge reported in Table 2.

Calibration of ClNO₂

ClNO₂ was calibrated by passing Cl₂ (50 ppmv in nitrogen, Air Liquide) over NaNO₂ (30 g) and NaCl (10 g) crystals in a glass flask (Thaler et al., 2011). The ClNO₂ generation efficiency was found to be improved by moistening the crystals by adding 2–3 drops of water. The gas mixture exiting the glass flask, which contains unreacted Cl₂, NO₂ and ClNO₂, was diluted in 5 slm air and sampled simultaneously with the CI-QMS and a cavity ring-down spectrometer that detects both NO₂ and ClNO₂ after thermal decomposition at 420 ^∘C to NO₂ (Thieser et al., 2016). ClNO₂ thermograms (Sobanski et al., 2016; Thieser et al., 2016) indicate that, under the flow and pressure conditions of these calibrations, ClNO₂ is thermally decomposed to NO₂ at 420 ^∘C (Reaction R20), but there is no significant loss at 200 ^∘C, the TDR temperature of the CI-QMS.

$$\begin{array}{}\text{(R20)}& {\displaystyle }{\mathrm{ClNO}}_{\mathrm{2}}+\mathrm{M}\left(\mathrm{420}{}^{\circ }C\right)\to \mathrm{Cl}+{\mathrm{NO}}_{\mathrm{2}}+\mathrm{M}\end{array}$$

The total uncertainty (≈25 %) associated with the calibration derives from uncertainty in the NO₂ cross section used to calculate NO₂ mixing ratios in the cavity ring-down spectrometer and the assumption that all ClNO₂ is detected as NO₂.

4.4 Hydrogen chloride

Using the discharge ion source we detect HCl as Cl⁻, ICl⁻ and I(CN)Cl⁻, presumably via Reactions (R21)–(23).

$$\begin{array}{}\text{(R21)}& {\displaystyle }& {\displaystyle }\mathrm{HCl}+\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}\to \mathrm{HCN}+\mathrm{I}\left(\mathrm{CN}\right){\mathrm{Cl}}^{-}(m/z\mathrm{188}\mathrm{and}\mathrm{190})\text{(R22)}& {\displaystyle }& {\displaystyle }\mathrm{HCl}+{\mathrm{ICN}}^{-}\to \mathrm{HCN}+{\mathrm{ICl}}^{-}(m/z\mathrm{162}\mathrm{and}\mathrm{164})\text{(R23)}& {\displaystyle }& {\displaystyle }\mathrm{HCl}+{\mathrm{CN}}^{-}\to \mathrm{HCN}+{\mathrm{Cl}}^{-}(m/z\mathrm{35}\mathrm{and}\mathrm{37}).\end{array}$$

Reactions (R22) and (R23) are weakly exothermic with reaction enthalpies of $\approx -\mathrm{25}$ and −73 kJ mol⁻¹, respectively. If available, the enthalpies of formation used to derive these values were taken from Goos et al. (2005), ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(HCl) $=-\mathrm{92}$, ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(HCN) =130, ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(Cl⁻) $=-\mathrm{234}$, ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(CN⁻) =61 kJ mol⁻¹, and Refaey and Franklin (1977), ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(ICl⁻) $=-\mathrm{155}$ kJ mol⁻¹. ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(ICN⁻) =92 kJ mol⁻¹ was calculated from ΔH${}_{\mathrm{f}}^{\mathrm{298}}$(ICN) =222 kJ mol⁻¹ (Goos et al., 2005) and the electron affinity of ICl (130 kJ mol⁻¹, (Miller et al., 2012). The $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ ion is stable in an aqueous solution (Chadwick et al., 1980), but a lack of thermodynamic data for it and for I(CN)Cl⁻ precludes calculation of the energetics of Reaction (R21).

As I(CN)Cl⁻, formed by reaction of dicyanoiodate anion with HCl in Reaction (R21), is the most abundant and specific product ion, HCl is generally monitored at m∕z 188. In the absence of interferences, the ratio of signals at m∕z 188 to m∕z 190 and m∕z 162 to m∕z 164 should be determined by the natural relative abundance of the ³⁵Cl and ³⁷Cl isotopes, which is ≈3.13. Plots of the relative ion signals at m∕z 188 versus m∕z 190 and m∕z 162 versus m∕z 164 obtained during the CYPHEX campaign are given in Fig. S5. The tight correlation and the slope of 3.09 for the ratio m∕z 162 to m∕z 164 is very close to the expected value, indicating that to a good approximation only one trace gas containing one Cl atom was measured. In contrast, the ratio of m∕z 188 to m∕z 190 is significantly lower than expected, and the correlation displays more scatter. This low ratio indicates that m∕z 190 suffers from interference from another trace gas. A likely candidate is HNO₃, detected as ${\mathrm{IHNO}}_{\mathrm{3}}^{-}$ at m∕z 190 (Lee et al., 2014). The signals at m∕z 188 and m∕z 162 are correlated very well (R²=0.96) indicating that they both represent HCl only, as no significant ClNO₂ was present during CYPHEX.

Calibration of HCl

A bottle of gaseous HCl diluted in N₂ (60 ppmv) was used to calibrate the CI-QMS during laboratory operation. The concentration of HCl was determined using UV-absorption spectroscopy (184.95 nm) using a cross section of $\mathrm{2.39}\times {\mathrm{10}}^{-\mathrm{19}}$ cm² molecule⁻¹ (Bahou et al., 2001), as described by Zimmermann et al. (2016). Once the sensitivity of the CI-QMS to HCl was established using bottled gas, the output of a laboratory-built permeation source was measured by comparing signals in the CI-QMS at m∕z 188. The permeation source consisted of a few mL of concentrated HCl-solution welded into a short length (4 cm) of 6.35 mm (OD) PFA tubing, housed in 20 cm of 12.7 mm (OD) PFA tubing (at 30 ^∘C) through which 50 sccm of air flows. The permeation rate measured was $\mathrm{5.2}\times {\mathrm{10}}^{-\mathrm{5}}$ sccm with an uncertainty of the HCl calibration of 10 %. The linearity of the CI-QMS signal with the HCl mixing ratio was characterized in the laboratory (R²=0.99) and is shown as Fig. S4b.

5 Dependence of sensitivity on relative humidity

The CI-QMS sensitivity for the trace gases discussed here is dependent on the amount of water vapour present in the IMR, which will vary with ambient relative humidity (RH). Broadly speaking, we observe a positive dependence of the sensitivity (see Fig. 5) on relative humidity between 0 % and 20 % at 25 ^∘C, with a flattening of the curve between 20 % and 80 % RH. This effect is generally explained by the reactions proceeding predominantly through clustered primary ions, e.g. I⁻(H₂O), which is observed at m∕z 145. Under weak de-clustering conditions, the product ions are also clustered with H₂O, confirming the participation of I⁻(H₂O). For the ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ primary ions used to detect SO₂ and HCl, the water clusters ${\mathrm{IO}}_{\mathrm{3}}^{-}$(H₂O)_n and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$(H₂O)_n are not observed or are very weak, even without de-clustering. We observe that the concentration of ${\mathrm{IO}}_{\mathrm{3}}^{-}$ and $\mathrm{I}(\mathrm{CN}{)}_{\mathrm{2}}^{-}$ in the primary-ion spectra is dependent on the presence of H₂O, which explains the RH dependence of the sensitivity of detection for SO₂ and HCl. All ambient measurements of the trace gases discussed here are therefore corrected for RH effects using calibration curves based on data such as those displayed in Fig. 5.

Figure 5 Dependence of the ion signal on the relative humidity (at 25 ^∘C and 1 bar) of the air sampled. For PAA + acetic acid on m∕z 59 only a combined humidity dependence is given as their contributions to the signal could not reliably be quantified.

6 Sensitivity, detection limits and total uncertainty

The total uncertainty of the measurement of any of the trace gases listed above is determined mainly by the uncertainty associated with the calibration method (and its reproducibility) but may also be influenced by, for example, scrubbing efficiency and drifts between background measurements (variable for different field campaigns). The response of the CI-QMS to any one trace gas may also vary over a period from days to a few weeks due to drifts in temperature, resolution of the mass spectrometer and degradation of the detector. The sensitivity (i.e. signal in Hz per pptv of trace gas) depends on the rate coefficient for the reaction between primary ions and trace gases and the yield of product ions. The sensitivity may also depend on relative humidity (abundance of H₂O clusters) and de-clustering potential (breakup of weak bonds). The limit of detection (LOD) is mainly dependent on variability in the background signal for the respective m∕z and can be calculated as 2 times the standard deviation when using synthetic (i.e. hydrocarbon-free) air or when passing the air through the scrubber (as usually performed during field measurements). In the text below (summarized in Table 2), we report sensitivities (in Hz pptv⁻¹) and limits of detection (LOD, 2σ, in pptv) obtained under typical measurement conditions (configuration iii from Sect. 3) and, when applicable, compare them to values obtained using ²¹⁰Po as an ion source.

6.1 PAN, PAA and acetic acid

When using ²¹⁰Po as an ion source, an LOD of 3 pptv PAN (1 s) is achievable, which is adequate for, e.g. airborne operation (Roiger et al., 2011) or flux measurements (Wolfe et al., 2009). The use of the RF discharge for PAN detection results in an increase in background signal (from a few hertz when using ²¹⁰Po to several hundred hertz when using the RF discharge ion source) even in hydrocarbon-free synthetic air. The LOD calculated from twice the standard deviation of a background measurement during the NOTOMO campaign is 34 pptv (1 s). The total uncertainty calculated from measurement precision, background subtraction (signal drifts, interpolation) and uncertainty in the calibration method is 15 % ±27 pptv. However, the uncertainty of the PAN measurement is highly dependent on the levels and variability of PAA and acetic acid present in the air, as their signal has to be interpolated and subtracted. In ambient air masses, the larger part of the signal at m∕z 59 with the RF discharge is due to acetic acid, which sometimes displays variability on the timescale of minutes. In this case the uncertainty of the background interpolation and therefore the overall uncertainty of the PAN measurement is drastically increased. This is essentially a selectivity problem, which limits deployment of the instrument for PAN measurements to more polluted regions where PAN mixing ratios regularly exceed 100 pptv and/or high time resolution is not necessary. As described above, the selectivity to PAN can be improved by switching off de-clustering (no acetic acid detection at m∕z 59), which comes with a significant reduction in sensitivity (see below).

In order to differentiate between PAA and acetic acid, the de-clustering voltage has to be modulated between 20 and 2 V. At the lower voltage, only PAA is detected but the resultant high chemical background and worsened sensitivity were found to lead to a poor limit of detection of a few hundred pptv (1 s) (see Table 2), which is about a factor of 100 higher than with use of ²¹⁰Po (4 pptv in 1 s). At the higher potential the LOD would be much better but the sensitivity to acetic acid at m∕z 59 reduces the selectivity of the measurement. The total uncertainty calculated from measurement precision, background subtraction (signal drifts, interpolation) and uncertainty in the calibration method is 20 %±39 pptv.

The LOD for acetic acid at m∕z 59 is 57 pptv (1 s) but the selectivity is reduced due to the PAA contribution. The total uncertainty calculated from measurement precision, background subtraction (signal drifts, interpolation) and uncertainty in the calibration method is 15 %±45 pptv.

6.2 SO₂

The sensitivity of the CI-QMS to SO₂ reported in Table 2 is dependent on the relative humidity (see Sect. 5) and is especially dependent on the de-clustering voltage, the best signal-to-noise ratio being found at 20 V. Although ${\mathrm{HSO}}_{\mathrm{4}}^{-}$ (m∕z 97) has the highest sensitivity of all product ions, we generally monitor the ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ ion (m∕z 207) as the background signal is lower and the detection limit improved. Figure S6 displays the correlation between both m∕z over a period of 4 weeks during the NOTOMO campaign. The correlation coefficient (R²=0.95) is large, from which we conclude that both m∕z can be used to calculate SO₂ mixing ratios. The detection limit for m∕z 207 is 56 pptv (1 s) (based on noise in background measurements during the NOTOMO field campaign), which is sufficient to monitor SO₂ in lightly polluted areas. At a lower temporal resolution and when monitoring only ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ and I⁻, the LOD can be improved to a few pptv (e.g. in 10 min). The total uncertainty, calculated from measurement precision, background subtraction (signal drifts, interpolation) and uncertainty in the calibration method is 20 %±23 pptv.

6.3 ClNO₂

Very good detection limits have been reported (Osthoff et al., 2008; Thornton et al., 2010; Phillips et al., 2012) for the measurement of ClNO₂ via I-CIMS using ²¹⁰Po ionization, a result of low background signal at m∕z 208 (${\mathrm{IClNO}}_{\mathrm{2}}^{-}$) and an efficient reaction with I⁻. Using ²¹⁰Po, Phillips et al. (2012) achieved an LOD (2σ) of 3 pptv (1 s), which can be compared to the value of 12 pptv (1 s) (see Table 2) obtained with the RF discharge ion source, the difference stemming from a higher chemical background signal. With an averaging interval of 5 min the LOD can be reduced to 2–3 pptv. ClNO₂ can also be detected as ICl⁻ (m∕z 162 and 164), which provides higher sensitivity compared to ${\mathrm{IClNO}}_{\mathrm{2}}^{-}$ (see Table 2) but can suffer from a significant interference due to HCl, which is likely to be present in air masses containing ClNO₂. For example, 1 ppbv HCl contributes a signal at m∕z 162 which is equivalent to 60 pptv ClNO₂ at this m∕z. Monitoring ClNO₂ at m∕z 208 is more specific, with an equivalent signal due to 1 ppb HCl of less than 10 pptv, which can be accounted for when measuring HCl in parallel (at m∕z 188; see above). It should be noted that the interference at m∕z 162 is not unique to the RF discharge but has also been observed when using a ²¹⁰Po ionizer (Phillips et al., 2012). The total uncertainty for ClNO₂ measurement, calculated from precision, background subtraction (signal drifts, interpolation) and uncertainty in the calibration method, is 30 %±6 pptv.

6.4 HCl

Sensitivities and product yields for several ions connected to HCl detection are reported in Table 2. As ICl(m∕z 162 and 164) suffers from a ClNO₂ interference (see above) and Cl⁻ (m∕z 35 and 37) could possibly arise from other Cl-containing species, the more specific ion I(CN)Cl⁻ (m∕z 188 and 190) is used to monitor HCl. The LOD for m∕z 188 is 135 pptv (1 s), which can be further improved by extended averaging if high time resolution is not required. The total uncertainty calculated from measurement precision, background subtraction (signal drifts, interpolation), scrubbing efficiency (it takes more time to remove HCl in the scrubber than, e.g. SO₂) and uncertainty in the calibration method is 20 %±72 pptv.

7 Application in the field

Our CI-QMS instrument has been deployed in different ground-based field campaigns including ones at coastal (CYPHEX, 2014), forested (IBAIRN, 2016) and mountain sites (NOTOMO, 2015) in Europe. In the following we present subsets of the data from these campaigns in order to indicate how the instrument with an RF discharge ion source (configuration iii in Sect. 3) performs in the field.

Figure 6 CI-QMS time series of SO₂ and HCl mixing ratios during the CYPHEX field campaign in Cyprus.

7.1 CYPHEX 2014

During CYPHEX (Cyprus Photochemistry Experiment, summer 2014), located at a coastal site on the eastern Mediterranean island of Cyprus, we measured chemically aged air masses with origins in continental Europe (Meusel et al., 2016; Derstroff et al., 2017). A time series of SO₂ and HCl for a 3-week period of the campaign is shown in Fig. 6. SO₂ was detected for the first time using the CI-QMS during CYPHEX in which observations of covariance between the signal at m∕z 207 and particulate sulfate provided the first clues to the identity of the mass peak as ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ and indications of sensitivity to SO₂. As we had not anticipated CI-QMS sensitivity to SO₂, calibration was performed post-campaign. We observed SO₂ mixing ratios as high as 11 ppbv, the plume-like nature of which strongly suggests nearby point sources such as ship traffic or air masses originating from power plants in continental Europe. Our measurements are consistent with other observations in the coastal Mediterranean boundary layer. Bardouki et al. (2003) found SO₂ mixing ratios up to 3 ppbv in Crete (August 2001) and Schembari et al. (2012) report average daily mean values of several ppbv in different western Mediterranean harbours measured in the summers of 2009 and 2010. Kanakidou et al. (2011) conclude that megacities can be hotspots of air pollution in the eastern Mediterranean, with average SO₂ mixing ratios of 1 ppbv measured in Crete (1997–1999), 8 ppbv in Istanbul (1998–2008), 10–15 ppbv in Athens (1995–1997) and 48 ppbv in Cairo (1999–2000), where about 70 % originates from industrial activities.

Similar to SO₂, the measurement of HCl was unexpected, the isotopic ratio of 3:1 for the signals at m∕z 162 and 164 (ICl⁻) providing evidence that the trace gas measured contained one Cl atom. The identification of HCl was confirmed during the campaign, in which a permeation source was built and used to periodically supply HCl to the CI-QMS. Calibration of the permeation source ensued post-campaign as described in Sect. 4.4. HCl mixing ratios up to 3 ppbv were observed, which could be attributed to the release of HCl from sea salt aerosols when polluted air masses from continental Europe reached the coastal site. The covariance between HCl and SO₂ in Fig. 6 suggests that an acid displacement mechanism involving SO₂ oxidation to H₂SO₄ and transfer of H₂SO₄ to aqueous sea salt aerosol was involved. The median value of the HCl mixing ratio for the CYPHEX campaign was 790 pptv, which can be compared with reports of median values of 600–700 pptv HCl over the Atlantic near Europe and values of up to 6 ppbv in the polluted coastal boundary layer on the Isles of Shoals, 10 km off the southern Maine coast, USA, with an average of 600 pptv (July–August 2004) (Keene et al., 2007, 2009).

Figure 7 CI-QMS time series of ClNO₂, SO₂, PAN mixing ratios and m∕z 59 during the NOTOMO field campaign in Germany. The signal at m∕z 59 was converted to ppbv, assuming that it consists only of acetic acid (no peracetic acid). The mixing ratios are therefore only an upper limit. For comparison with the PARADE campaign at the same location see Fig. S7.

7.2 NOTOMO 2015

The mountain-site campaign NOTOMO (NOcturnal chemistry at the Taunus Observatory: insights into Mechanisms of Oxidation, summer 2015) took place in a rural location in south-western Germany with significant urban influence (Sobanski et al., 2017). A time series of SO₂, ClNO₂ and PAN mixing ratios and the signal at m∕z 59 (contributions from acetic and peracetic acid) is displayed in Fig. 7. During NOTOMO, SO₂ was monitored as ISO₃- (m∕z 207) and HSO₄- (m∕z 97), and a very good correlation between both signals (R²=0.95) within a period of 4 weeks confirmed that both ions reliably represent the same molecule (see Fig. S6). The SO₂ mixing ratios exceeded 1 ppbv on most days, with maximum values up to 5 ppbv. The likely origins of SO₂ at this site are emissions from coal-burning power plants in the local Rhine–Main urban conglomeration and the heavily industrialized Ruhr area to the north-west.

ClNO₂ was detected during NOTOMO as ${\mathrm{IClNO}}_{\mathrm{2}}^{-}$ (m∕z 208 and 210). Mixing ratios ranged from 0 to 500 pptv and were above 50 pptv during 10 out of 29 campaign nights. High levels of ClNO₂ were generally associated with mixed marine and continental air masses from the north-west which had passed over the English Channel and the polluted Ruhr area. The data are consistent with previous measurements (using the CI-QMS equipped with a ²¹⁰Po ionizer) at the same location and similar time of year (Phillips et al., 2012), whereby comparable ClNO₂ mixing ratios were observed (see Fig. S7).

We measured PAN at m∕z 59 and observed mixing ratios generally ranging between 0 and 2 ppbv throughout NOTOMO, occasionally reaching 3 ppbv. PAN levels predominantly peaked in the afternoon, when photochemical activity is usually highest. Compared with results from PARADE (see Fig. S7) in which PAN had been measured with a precursor version of this instrument with a ²¹⁰Po ionizer, frequency and amplitude of the PAN mixing ratios throughout the campaign were very similar. However, due to the high and variable chemical background at m∕z 59, as already pointed out in Sect. 4.1, the detection limit during NOTOMO was about an order of magnitude worse than in PARADE (see Table 2). As we did not require higher temporal resolution than a few minutes for analysis, this was not a key issue here. The measurements with the RF discharge ion source had adequate sensitivity towards PAN in this moderately polluted region and have the additional advantage over ²¹⁰Po of simultaneous detection of SO₂ and HCl. In contrast, the differentiation between PAA and acetic acid is problematic. During NOTOMO, we were unaware of the sensitivity towards acetic acid at m∕z 59, and the signal without PAN (i.e. during titration of CH₃C(O)O₂ with NO) was measured with 20 V de-clustering only and therefore represents a combined signal of PAA and acetic acid. Although peracetic acid calibration was performed during the campaign, this is not considered reliable because the PAA diffusion source also contains significant amounts of acetic acid. For this reason, we present only an upper limit for acetic acid, ranging between 0 and 8 ppbv. From the PAN-to-PAA ratio calculated for PARADE (about 10) we would also expect several hundred pptv of PAA to be present (see Fig. S7), which would lower this approximate acetic acid mixing ratio significantly.

Figure 8 CI-QMS time series of SO₂, HCl, PAN, PAA mixing ratios and m∕z 59 during the IBAIRN campaign, which was undertaken in boreal forest. The red SO₂ trace was obtained using a TEI 43 CTL fluorescence analyser (SMEAR II). The signal at m∕z 59 was converted to ppbv assuming that it consists only of acetic acid (no peracetic acid). The mixing ratios are therefore only an upper limit.

7.3 IBAIRN 2016

The IBAIRN campaign (Influence of Biosphere–Atmosphere Interactions on the Reactive Nitrogen budget, summer 2016) took place in boreal forest at Hyytiälä, Finland, an area with large biogenic emissions and low-NO_x conditions (Liebmann et al., 2018). The CI-QMS inlet was located at about 6 m height within the canopy. A time series of SO₂, HCl, PAN, PAA and the signal at m∕z 59 that has contributions from both acetic and peracetic acid is shown in Fig. 8.

During IBAIRN SO₂ was monitored by the CI-QMS as ${\mathrm{ISO}}_{\mathrm{3}}^{-}$ (m∕z 207) with mixing ratios up to 1 ppbv. SO₂ mixing ratios were largest when the air originated from the north-east (point sources like coal-burning power plants in northern Finland and Russia) but only occasionally exceeded 100 pptv in this remote, forested environment. Independent SO₂ measurements, using a TEI 43 CTL fluorescence analyser (SMEAR II station, University of Helsinki), were made on a tower at a 16 m height (5 m distant from the CI-QMS inlet but approximately at canopy height) and allow a direct comparison to be made. The datasets are generally in good agreement (Fig. 8a), although some SO₂ plumes were only observed at the higher inlet due to strong gradients in trace gas concentrations resulting from boundary layer dynamics.

HCl (Fig. 8b) was measured as I(CN)Cl⁻ (m∕z 188 and 190) with mixing ratios up to 300 pptv showing a distinct diurnal profile with a maximum in the afternoon, which reflects temperature-dependent changes in partitioning between the gas phase and particle phase in which HCl is converted into NH₄Cl.

The combined PAA and acetic acid signal at m∕z 59 obtained with de-clustering at 20 V (Fig. 8e) displays the diurnal profile expected from photochemically generated trace gases with night-time depositional losses. However, as explained in Sect. 7.2, we cannot easily separate the contribution to the signal from PAA and acetic acid. When NO is added (see Fig. S3), the signal at m∕z 59 with low de-clustering is due to PAA only (Fig. 8d), but the PAA detection limit is poor due to a low sensitivity and elevated and variable background signal during zeroing. The PAA contribution to the total signal at m∕z 59 is likely to be substantial but cannot be calculated as the sensitivity of the CI-QMS to PAA during de-clustering (Fig. 8e) is unknown, a result of the presence of unknown amounts of acetic acid in the PAA diffusion source. The use of a ²¹⁰Po ionizer would have greatly improved the PAN (see Fig. 8c) and PAA data quality, as evidenced by the PAN and PAA measurements reported for the same location during 2010 (Phillips et al., 2013; Crowley et al., 2018).

8 Conclusions

The CI-QMS with an RF discharge ion source is a promising alternative to similar instruments using ²¹⁰Po-based ion sources and can be deployed in environments for which permission to use ²¹⁰Po may be difficult or impossible to obtain or to which transportation of ²¹⁰Po is not feasible. The use of the RF discharge results in an extension of the established detection schemes (e.g. for ClNO₂ and PAN) using I⁻ ions, and we have identified ion schemes involving ${\mathrm{IO}}_x^{-}$ and $\mathrm{I}(\mathrm{CN}{)}_x^{-}$ primary ions that additionally enable detection of SO₂ and HCl. Detection limits (2σ, 1 s integration) are 56 pptv for SO₂, 135 pptv for HCl and 12 pptv for ClNO₂, which makes the CI-QMS a useful tool for investigation of atmospheric processes related to sulfur and chlorine chemistry. Application of the instrument with the RF discharge ion source for PAN detection is limited to polluted environments where mixing ratios usually exceed a hundred pptv and a high temporal resolution is not needed. This restriction is mainly due to a background count rate that is higher by 1–2 orders of magnitude compared to the use of ²¹⁰Po and due to a strong dependence of the measurement uncertainty on the variability of the subtracted interpolated background signal (consisting of PAA and acetic acid). A PAN detection limit (in the absence of PAA and acetic acid) of 34 pptv (1 s) was obtained, though this value will rarely be reached in boundary layer air masses where acetic acid and peracetic acid are abundant. Similarly, while sensitive detection of PAA (requiring de-clustering) is precluded by the detection of acetic acid at the same m∕z, the selective detection of acetic acid is uncertain due to the contribution of PAA.

The deployment of the CI-QMS with RF discharge and its advantages and disadvantages compared to instruments using ²¹⁰Po-based ionization are illustrated in three campaign datasets, which demonstrate its potential to monitor trace gases at mixing ratios ranging from a few tens of pptv to a few ppbv. If the scientific focus is on PAA and PAN, the RF-discharge-based CI-QMS is clearly disadvantaged compared to the more selective and sensitive ²¹⁰Po-based ionization. On the other hand, the potential to measure ClNO₂ without logistical obstacles related to transport and (mobile) operation of radioactive sources and the added benefit of simultaneous measurement of HCl and SO₂ may in some instances tip the balance in its favour.