Iodide CIMS and m/z 62: the detection of HNO 3 as NO 3 − in the presence of PAN, peroxyacetic acid and ozone

. Chemical ionisation mass spectrometry (CIMS) using I − (the iodide anion), hereafter I-CIMS, as a primary reactant ion has previously been used to measure NO 3 and N 2 O 5 both in laboratory and ﬁeld experiments. We show that reports of large daytime mixing ratios of NO 3 and N 2 O 5 (both usually present in detectable amounts only at night) are likely to be heavily biased by the ubiquitous presence of HNO 3 in the troposphere and lower stratosphere. We demon-strate in a series of laboratory experiments that the CIMS detection of HNO 3 at m/z 62 using I − ions is efﬁcient in the presence of peroxy acetyl nitric anhydride (PAN) or peroxyacetic acid (PAA) and especially O 3 . We have characterised the dependence of the sensitivity to HNO 3 detection on the presence of acetate anions (CH 3 CO 2 − , m/z 59, from either PAN or PAA). The loss of CH 3 CO 2 − via conversion to NO 3 − in the presence of HNO 3 may represent a signiﬁcant bias in I-CIMS measurements of PAN and PAA in which continuous calibration (e.g. via addition of isotopically labelled PAN) is not carried out. The greatest sensitivity to HNO 3 at m/z 62 is achieved in the presence of ambient levels of O 3 whereby the thermodynamically disfavoured, direct reaction of I − with HNO 3 to form NO 3 − is bypassed by the formation of IO x − , which reacts with HNO 3 to form, for ex-ample, iodic acid and NO 3 − . The ozone and humidity dependence of the detection of HNO 3 at m/z 62 was characterised in laboratory experiments and applied to daytime, airborne measurements in which good agreement with measurements of the I − (HNO 3 ) cluster ion (speciﬁc for HNO 3 detection) was obtained. At high ozone mixing ratios, we show that the concentration of I − ions in our ion–molecule reactor (IMR) is signiﬁcantly depleted. This is not reﬂected by changes in the measured I − signal at m/z 127 as the IO x − formed does not survive passage through the instrument but is likely detected after fragmentation to I − . This may result in a bias in measurements of trace gases using I-CIMS in stratospheric air masses unless a calibration gas is continuously added or the impact of O 3 on sensitivity is characterised.


Introduction
The use of iodide anions (I − ) as primary ions in massspectrometric studies of ion-molecule reactions has a long history. Fehsenfeld et al. (1975) and Davidson et al. (1978) established that the nitrate anion (NO 3 − , m/z 62) was formed in a rapid reaction between I − and N 2 O 5 . NO 3 − was also identified as the main product of the reaction between I − and ClONO 2 (Huey et al., 1995). The large rate constants for reaction of I − with N 2 O 5 and ClONO 2 led to the development of chemical ionisation mass spectrometry (CIMS) using I − primary ions (henceforth I-CIMS) in kinetic studies of heterogeneous, atmospheric reactions (e.g. Hanson and Ravishankara, 1991) and more recently I-CIMS has found widespread deployment for measurement of atmospheric trace gases (Huey, 2007, and references therein). Early field measurements utilised I-CIMS to detect N 2 O 5 and peroxyacetyl nitric anhydride (PAN, CH 3 C(O)O 2 NO 2 ) (Slusher et al., 2004), but since then the range of molecules that have been detected using I − has greatly increased and trace gases as diverse as inorganic radicals and halogenates and a host of organic species are now routinely measured (Huey, 2007;Lee et al., 2014;Iyer et al., 2017;Riva et al., 2019). In this work, we focus on the detection of two atmospherically important trace gases N 2 O 5 and HNO 3 using a CIMS operating with I − reactant ions. R. Dörich et al.: Detection of HNO 3 at m/z 62 using iodide CIMS Both N 2 O 5 and HNO 3 are formed in the atmosphere by the sequential oxidation of NO, which has both anthropogenic and natural sources. In a well-known series of reactions (Lightfoot et al., 1992;Atkinson et al., 2004), NO is oxidised (Reactions R1 and R2) by reaction with O 3 or peroxyl radicals (RO 2 ) to NO 2 , which during the day, may be removed by reaction with OH to form HNO 3 (Reaction R3) and during the night to form N 2 O 5 (Reactions R4 and R5).
Both HNO 3 and N 2 O 5 have important, non-gas-phase loss processes (Crowley et al., 2010) such as uptake to particles and other surfaces. In addition, N 2 O 5 can thermally dissociate back to NO 3 . The chain of reactions to form N 2 O 5 is broken during the day as NO 3 is generally photolysed within a few seconds (Wayne et al., 1991) and also reacts with NO so that N 2 O 5 is expected to be present at significant levels only at night-time.
The detection of N 2 O 5 using I − reactant ions can be achieved by monitoring either the NO 3 − product at m/z 62 (see above) or the adduct ion at m/z 235 (Kercher et al., 2009). The former is reported to be more sensitive and less dependent on water vapour concentrations but less specific, with large and highly variable background signals potentially arising from trace gases such as NO 3 , ClONO 2 and BrONO 2 . Despite this, night-time N 2 O 5 has been monitored in ambient air (as NO 3 − ) using I − reactant ions, showing reasonable agreement with optical methods (Slusher et al., 2004;Dubé et al., 2006;Chang et al., 2011).
During a recent, airborne deployment of our I-CIMS, we monitored NO 3 − at m/z 62 in an attempt to detect N 2 O 5 during two night-time flights. The air masses we investigated were mainly in the tropical free and upper troposphere and lower stratosphere, and we did not expect significant interference from, for example, halogen nitrates at m/z 62. However, our airborne measurements (described in detail in Sect. 4) revealed a large and variable signal at m/z 62 during both the day and night. To illustrate this, raw signals obtained during daytime when the aircraft sampled air masses with varying degrees of stratospheric influence are displayed in Fig. S1 in the Supplement. The signal at m/z 62 is large and highly variable and is not affected by addition of NO to the heated inlet, ruling out its assignment to either N 2 O 5 or NO 3 (see below). The great increase in signal when entering the lower stratosphere and the obvious correlation with O 3 (Fischer et al., 1997;Popp et al., 2009) provided an early clue to the identity of the trace gas detected at m/z 62 which we initially assigned to HNO 3 . Our results thus appeared to contrast the conclusions of a previous observation of a large daytime signal at m/z 62 when deploying an I-CIMS (in this case in the boundary layer), which was interpreted as resulting (at least in part) from high levels of daytime NO 3 and/or N 2 O 5 (Wang et al., 2014). Based on complementary laboratory experiments, Wang et al. (2014) showed, in accord with earlier investigations (Fehsenfeld et al., 1975;Huey et al., 1995), that HNO 3 is not detected sensitively at m/z 62 using I-CIMS.
The unexpected observation of a large daytime signal at m/z 62 during airborne operation led us to perform a series of laboratory experiments to identify potential "interfering" trace gases at this mass-to-charge ratio when using I-CIMS. In contrast to the conclusions drawn from previous studies, our laboratory and airborne measurements show that, during daytime, the predominant contributor to m/z 62 when sampling ambient air (in the presence of ozone) is likely to be HNO 3 .

Experimental details
The I-CIMS we used in our laboratory and airborne investigations (see Fig. 1) is similar to that described by Slusher et al. (2004) and Zheng et al. (2011) and was originally constructed in collaboration with Georgia Tech as a prototype instrument of the company THS (http://thsinstruments.com, last access: 30 July 2021). It is essentially a hybrid of the instruments described by Phillips et al. (2013) and Eger et al. (2019), the former using a 210 Po ion source, the latter an electrical discharge source but with improved (digital) control of the MS settings enabling different mass-to-charge ratios to be monitored using different potentials for the collisional dissociation of cluster ions. For all the experiments described below, the 210 Po ion source was used to generate I − as this configuration has much better sensitivity for PAN, the main target trace gas during the deployment of the I-CIMS on the HALO aircraft (High Altitude Long range platform for atmospheric Observations). The set-up for PAN detection includes a heated inlet section (∼ 170 • C, 100 mbar, residence time ∼ 40 ms) to thermally dissociate PAN to CH 3 C(O)O 2 which subsequently reacts with I − to form the acetate anion (CH 3 CO 2 − ) which is detected at m/z 59. At this inlet temperature and pressure, the lifetime of PAN with respect to thermal decomposition < 3 ms (Ammann et al., 2021). For N 2 O 5 , the lifetime with respect to its thermal dissociation to NO 2 and NO 3 is ∼ 0.5 ms (IUPAC, 2021) so that N 2 O 5 is stoichiometrically converted to NO 3 and the instrument measures the sum of N 2 O 5 and NO 3 at m/z 62. In order to separate PAN signals from those of peroxyacetic acid (CH 3 C(O)OOH, also detected as CH 3 CO 2 − at m/z 59), we periodically add NO (∼ 5 × 10 12 molecule cm −3 ) to the inlet to remove CH 3 C(O)O 2 and thus eliminate sensitivity to PAN. As NO reacts more rapidly with NO 3 than with CH 3 C(O)O 2 at 170 • C (k NO+NO 3 = 2.3 × 10 −11 cm 3 molecule −1 s −1 , k NO+CH 3 C(O)O 2 = 1.4 × 10 −11 cm 3 molecule −1 s −1 ; IUPAC, Figure 1. Schematic diagram illustrating the central components of the I-CIMS used in this work. IMR = ion-molecule reactor, CDC = collisional dissociation chamber, F 1 = 1250 cm 3 (STP) min −1 , F 2 = 50 cm 3 (STP) min −1 , F 3 = 750 cm 3 (STP) min −1 , F 4 = 4 cm 3 (STP) min −1 , F 5 = 50 cm 3 (STP) min −1 . p1 = 100 mbar, p2 = 24 mbar, p3 = 0.6 mbar, p4 = 6 × 10 −3 mbar, p5 = 9 × 10 −5 mbar. The heated inlet is made of PFA tubing. TGI = Trace-gas inlet. When overflowing the inlet line, the valve to the exhaust line is closed. When sampling air, the valve to the N 2 bottle is closed. 2021), the concentration of NO added is also sufficient to quantitatively titrate NO 3 to NO 2 and thus provides a measure of the "background" signal at m/z 62 in the absence of NO 3 and N 2 O 5 .
During airborne operation on HALO, the dynamic pressure generated in a forward facing trace gas inlet (TGI) located on top of the aircraft (see Fig. 1) was used to create a flow of air through 1/4 in. (OD) PFA tubing sampling at an angle of 90 • to the flight direction. The 1/4 in. tubing was attached to a 1/2 in. (OD) PFA tube attached to an exhaust plate at the underside of the aircraft to create a fast "bypass" flow. The bypass flow was sub-sampled (again at 90 • and by 1/4 in. PFA tubing heated to 40 • C) by the 1.4 L (STP) min −1 flow into the I-CIMS. Sub-sampling twice at 90 • to the flow was helpful in reducing the number of large particles (e.g. cloud droplets) that could enter the thermal dissociation inlet and IMR.
The thermal dissociation inlet of the I-CIMS is regulated to a pressure of 100 mbar, which results in a pressure in the ion-molecule reactor of 24 mbar. This way, a stable pressure in the thermal dissociation inlet and the ion-molecule reactor (IMR) was maintained at altitudes up to ∼ 15 km. Prior to take off, the inlet line and TGI were flushed with nitrogen to prevent contamination by the high levels of pollutant trace gases at the airport. As described in Eger et al. (2019) negative ions exiting the IMR were declustered in passage through a collisional dissociation region (CDC, 0.6 mbar) before passing through an octopole ion guide (6 × 10 −3 mbar) and a quadrupole for mass selection (9 × 10 −5 mbar) prior to detection using a channeltron. I − ions were generated by combining flows of 4 cm 3 (STP) min −1 CH 3 I/N 2 (400 ppmv) with 750 cm 3 (STP) min −1 N 2 and passing the mixture through a 370 MBq 210 Po source. Under standard operating conditions (including airborne deployment), a constant amount of H 2 O was added to the IMR by flowing 50 cm 3 (STP) min −1 N 2 (at 1 bar pressure) through a 30 cm length of waterpermeable 1/8 in. tubing (Permapure) immersed in water. The 50 cm 3 (STP) min −1 flow of N 2 acquires a relative humidity close to 100 % in transit through the tubing and is subsequently mixed with the dry N 2 flow and sample air. Under these conditions, the ratio of signals at m/z 145 (I − (H 2 O)) to that at m/z 127 (I − ) was 0.068. By comparison with calibration curves (see Fig. S2 and associated text) this indicates an H 2 O concentration in the IMR of ∼ 4 × 10 14 molecule cm −3 . For laboratory tests, the amount of water in the IMR could be increased by reducing the pressure in the permeable tube (thus increasing the mole fraction of H 2 O) or set to zero by bypassing the humidifier.
Based on a (calculated) literature value for the free energy of formation of I − (H 2 O) 1 of −6.1 kcal mol −1 (Teiwes et al., 2019), we derive an equilibrium constant (at 298 K) of K 6 = 1.16 × 10 −15 cm 3 molecule −1 for the formation and thermal dissociation of I − (H 2 O) 1 : With an H 2 O concentration (in the IMR) of 3.9 × 10 14 molecule cm −3 , this implies that the ratio [I − (H 2 O) 1 ]/[I − ] = 0.45. Our measured ratio of signals at m/z 145 (I − (H 2 O)) : m/z 127 (I − ) was a factor ∼ 6 lower, reflecting the fact that, even when the declustering potential is reduced to its minimum value, most I − (H 2 O) ions do not survive the CDC region.
During extended operation of the CIMS, changes in sensitivity were captured by monitoring the primary ion signal (I − and its water cluster). Background signals at each of the mass-to-charge ratios monitored were obtained by passing the sampled air through a tubular scrubber (aluminium) filled with stainless-steel wool heated to 120 • C.
3 Laboratory characterisation 3.1 Detection of HNO 3 at m/z 62: the role of ozone As described above, our observations of a clear correlation between m/z 62 and O 3 mixing ratios during the first HALO deployment of the I-CIMS strongly suggested that HNO 3 was the origin of the signal although previous experiments had shown that I − does not react with HNO 3 to form NO 3 − . In order to determine the sensitivity of our I-CIMS to HNO 3 , we constructed a permeation source in which a 20 cm 3 (STP) min −1 flow of zero air was passed through a 1 m length of PFA tubing (0.125 in. OD) which was formed into a coil and submerged in an aqueous solution of 65 % HNO 3 held at 50 • C. The permeation rate was determined by passing the 20 cm 3 (STP) min −1 flow through an optical absorption cell and measuring the optical extinction at 185 nm where the absorption cross-section of HNO 3 is well known (Dulitz et al., 2018). For the I-CIMS calibration, the 20 cm 3 (STP) min −1 output was dynamically diluted to generate a mixing ratio of between 5 and 50 ppbv. Based on uncertainties in the absorption cross-section (5 %), the reproducibility of the optical measurement and the dilution factor, the uncertainty of the HNO 3 mixing ratio is estimated as 15 %. Figure 2 shows the response of the I-CIMS at m/z 62 to addition of various amounts of HNO 3 . Throughout the paper, when presenting raw data, we generally normalise the I-CIMS signal by dividing by the primary ion signal at m/z 127. This is standard practice and corrects for drifts in the CH 3 I flow which may occur over several hours after the instrument was switched from standby mode to operational mode. It also accounts for longer-term drifts caused by the weakening activity of the 210 Po source over the duration of a measurement campaign (months) or since the last calibration and for loss of detector sensitivity over similar time periods. We show later that, when adding large concentrations of reactants that significantly deplete the primary ion signal at m/z 127, this procedure may lead to bias is some measurements. For this reason, when detecting HNO 3 at m/z 190 (see later), we normalise to an interpolated signal at m/z 127 that was measured when the air was scrubbed.
In a further series of experiments, we measured the response of the I-CIMS to HNO 3 when adding O 3 to the zero air. The results, also plotted in Fig. 2 (black symbols), indicate a factor ∼ 250 increase in the signal at m/z 62 when ∼ 500 ppbv ozone was added. There are two possible explanations for this observation. The first involves conversion of NO 2 impurity (which is present as a ∼ 8 % impurity in the HNO 3 permeation flow) to NO 3 and N 2 O 5 (Reactions R1, R4, and R5), which are subsequently detected. This can however be ruled out as the rate-limiting step in the formation of NO 3 is the slow reaction between NO 2 and O 3 with k 4 = 3.5 × 10 −17 cm 3 molecule −1 s −1 at room temperature (Atkinson et al., 2004). The addition of 1000 ppbv O 3 (equivalent to a concentration of 2.4 × 10 13 molecule cm −3 ) would only convert an insignificant fraction of the NO 2 to NO 3 in the ∼ 40 ms reaction time available from the point of mixing to the IMR. This was confirmed by adding NO (7.7 ppm) to the inlet, which would remove any NO 3 (see above) and observing no change in the signal at m/z 62.
The second explanation is that the presence of O 3 results in the generation of further reagent ions that can react with HNO 3 . Iodide anions are known to react with O 3 , leading, in a series of exothermic reactions, to the formation of iodate (Williams et al., 2002;Teiwes et al., 2018;Bhujel et al., 2020).
In this scheme, Reaction (R8) is rate-limiting (k 8 ∼ 1 × 10 −12 cm 3 molecule −1 s −1 ; Bhujel et al., 2020), whereas the further steps (Reactions R9-R10) in the sequential oxidation to iodate proceed with rate constants at least 2 orders of magnitude larger (Teiwes et al., 2018;Bhujel et al., 2020). IO − and IO 2 − also react with O 2 to reform O 3 : with rate coefficients of k 11 = 3.2 × 10 −14 cm 3 molecule −1 s −1 and k 12 = 1.3 × 10 −14 cm 3 molecule −1 s −1 (Bhujel et al., 2020). With the O 3 (∼ 1-5 ×10 10 molecule cm −3 ), O 2 concentrations (∼ 4× that of O 3 ) and reaction times used in their studies (Teiwes et al., 2018(Teiwes et al., , 2019Bhujel et al., 2020), IO 3 − was observed to be the dominant form of IO x − . In the presence of water vapour, I − is also present as a hydrate I − (H 2 O) (see above) for which, according to Teiwes et al. (2019), the rate coefficient for reaction with O 3 (Reactions R13a, R13b) is a factor ∼ 40 larger than k 8 and results in the formation of IO 2 − and I − : As Reaction (R8) is rate-limiting, this implies an increase in the amount of, for example, IO 3 − formed in the IMR in the presence of water. In most regions of the troposphere and lower atmosphere, ozone mixing ratios lie between 30 and > 1000 ppbv. An ambient ozone concentration of 50 ppbv results in a concentration in the IMR of > 10 10 molecules cm −3 . The large rate coefficients for Reactions (R9) and (R10) and the reactions of IO − and IO 2 − with O 2 result in the rapid inter-conversion of I − , IO − , IO 2 − and IO 3 − , which results (for a given RH and ozone concentration) in a quasi-equilibrium between IO x − anions. We explored the relevance of these reactions for our I-CIMS by carrying out a set of experiments in which varying amounts of O 3 were added to the inlet and the mass-to-charge ratios corresponding to IO − (m/z 143), IO 2 − (m/z 159) and IO 3 − (m/z 175) were monitored; the results are depicted in Fig. 3.
First, we note that all three mass-to-charge ratios were indeed observed, but only under conditions where the CDC potential was set to the lowest value at which ions still reach the detector. The dependence of the various IO x − anions on the O 3 mixing ratio is broadly as expected from the reaction scheme (Reactions R8-R12) listed above: the major contributor to IO x − at low [O 3 ] is IO − , which is converted to IO 2 − and IO 3 − more efficiently as O 3 increases, while the total concentration of IO x − increases approximately linearly. At the maximum O 3 mixing ratio used (577 ppbv) there is (following dilution) 375 ppbv in the IMR, which translates to a concentration (at 24 mbar and ∼ 298 K) of 2.1 × 10 11 molecule cm −3 . This O 3 concentration is comparable with those used by Teiwes et al. (2018) (∼ 1-4 ×10 11 molecule cm −3 ) or Bhujel et al. (2020) (∼ 4 × 10 10 molecule cm −3 ) in their ion-trap-based, kinetic investigations of the formation of IO x − when reacting I − with O 3 . Their observation that IO 3 − is the dominant anion is however not consistent with our results, which indicate that IO 3 − represents only ∼ 35 % of the total IO x − signal. The relative abundance of each IO x − depends not only on the O 3 concentration but also on the reaction time, which, for both Teiwes et al. (2018) and Bhujel et al. (2020) was between 10-100 ms. Based on the flow into the IMR, its volume (∼ 50 cm 3 ) and the pressure, we calculate a similar residence time (for neutrals) of about 25 ms. We also considered the possibility that the application in our I-CIMS of a potential difference between the entrance and exit of the IMR (to optimise ion transmission) could result in a significantly shorter IMR residence time for ions. This was assessed by calculating the drift velocity (V d ) in the IMR from the electric field strength (E ∼ 12 V m −1 ) and the ion mobility (µ).
R. Dörich et al.: Detection of HNO 3 at m/z 62 using iodide CIMS The electrical mobility of I − was calculated for our conditions (using the Mason-Schamp equation) as ∼ 0.15 m 2 V −1 s −1 using a collision cross-section (for an I − /N 2 pair) of 9 × 10 −16 cm 2 molecule −1 (McCracken, 1952). Via Eq. (1), this results in a drift velocity of 1.8 m s −1 , or an ion residence time (in the ∼ 8 cm long IMR) of 44 ms, which is comparable to the residence time of neutrals. Note that the IMR reaction times we derive are only approximate as we do not take into account the mixing and flow dynamics in the IMR, which are likely to be complex (and possibly shorter than 25 ms) owing to sampling via a critical orifice. While we cannot rule out that our observation of IO − (and not IO 3 − ) being the dominant ion signal is partially caused by differences in reaction times, slight differences in O 3 concentrations and differences in temperature (our IMR is at ∼ 15 • C above ambient temperature owing to the heated inlet), we note that the higher pressure of O 2 (factor ∼ 1-10 ×10 4 ) in our IMR is likely to have a large effect. The presence of O 2 converts IO 3 − back to IO − , thus competing with further oxidation (via reaction with O 3 ) to IO 3 − . Additionally, the high IMR pressures (24 mbar) in our experiments are ∼ 6 orders of magnitude higher than the ∼ 10 −5 mbar available in the ion-trap experiments of Teiwes et al. (2018) and Bhujel et al. (2020), which will result in more rapid thermalisation of the ions present and prevent potentially non-thermal reactions and thus bias to the rate coefficients derived.
The effect of adding H 2 O to the IMR was explored in a further set of experiments, and the variations of the signals at mass-to-charge ratios corresponding to IO x − with [H 2 O] are displayed in Fig. 4. The experiments were carried out with the O 3 mixing ratio fixed at either 70 or 120 ppbv, close to that typically found in the lower troposphere (∼ 20-100 ppbv). At the lowest H 2 O concentrations in our experiments, the total IO x − signal is about 30 counts. This increases by a factor of ∼ 10 when [H 2 O] IMR increases to 3×10 15 molecule cm −3 . Increasing the O 3 mixing ratio from 70 to 120 ppbv results in an increase in the signals at m/z 175 (IO 3 − ) and m/z 159 (IO 2 − ) at all water vapour concentrations, whereas the signal at m/z 143 (IO − ) is reduced at the lowest water vapour concentrations. These observations reinforce the concept of a larger rate coefficient for reaction of I − (H 2 O) with O 3 (Reaction R13a) compared to I − (Reaction R8) (Teiwes et al., 2019) and the sequential conversion of IO − to more oxidised forms as described by Reactions (R8)-(R10).
Having established that all of the expected IO x − anions are present in our IMR, we can propose a route for HNO 3 detection as NO 3 − , which involves transfer of a proton from HNO 3 (a very strong acid) to the conjugate base of the respective iodine containing acids (hypoiodous, iodous and iodic acid):   (Goos et al., 2005), we calculate that Reaction (R16) is exothermic by ∼ 62 kJ mol −1 .
As described above, the O 3 dependence of the ion signals we observe for IO − , IO 2 − and IO 3 − is consistent with the sequential oxidation of I − by O 3 . However, the relative ion abundance we observe at the detector does not necessarily reflect the relative concentration of the ions in the IMR, and we cannot assign the individual contribution of any single IO x − anion to HNO 3 detection. We are unable to completely shut of collisional dissociation in our I-CIMS, which may be a characteristic that is peculiar to our instrument as we do not detect weakly bound I − (R(O)OH) clusters which are commonly monitored in other instruments utilising I − chemical ionisation (Lee et al., 2014). Hence, our relative sensitivity to the IO x − components is unknown. In order to confirm that IO x − is responsible for detection of HNO 3 , we examined the depletion of the signals at m/z 143, m/z 159 and m/z 175 when adding very large concentrations of HNO 3 to the IMR. The results, summarised in Fig. 5, indicate that all three IO x − ions are removed when the HNO 3 mixing ratio was increased from zero to 80 ppbv, but with different fractional changes. This can be understood if, for example, the individual IO x − reacts with HNO 3 with different rate coefficients. The solid lines in Fig. 5 represent exponential decays of each ion, with rate coefficients of ∼ 10 × 10 −10 cm 3 molecule −1 s −1 for HNO 3 + IO 3 − , ∼ 7 × 10 −10 cm 3 molecule −1 s −1 for HNO 3 + IO 2 − and ∼ 3 × 10 −10 cm 3 molecule −1 s −1 for HNO 3 + IO − . These approximate values were derived by converting the HNO 3 mixing ratio into a concentration in the IMR and assuming pseudofirst-order behaviour (i.e. negligible depletion of HNO 3 ) so that (using IO 3 − as example) where S(IO 3 − ) t and S(IO 3 − ) 0 are the signals at m/z 175 after and prior to addition of HNO 3 , respectively. [HNO 3 ] IMR is the concentration (molecule cm −3 ) of HNO 3 in the IMR, k (cm 3 molecule −1 s −1 ) is the rate coefficient for reaction between HNO 3 and IO 3 − , and t is the reaction time, which we assume to be 25 ms (see above). This analysis assumes that the re-establishment of equilibria between IO x − is minimal on the timescale of the reaction between any single IO x − and HNO 3 . The results indicate qualitatively that IO 3 − is the most reactive of the IO x − anions towards HNO 3 but that all three contribute to HNO 3 detection. The depletion of the summed IO x − signals versus the accompanying increase in signal due to NO 3 − at m/z 62 is displayed in Fig. 6, which indicates a roughly linear relationship, confirming that IO x − is mainly responsible for detection of HNO 3 in our I-CIMS. We note that the absolute increase in signal at m/z 62 is about a factor of 100 greater than the reduction in the signal from IO x − , implying that the detection of IO 3 − in our instrument is inefficient.
While the reactions of IO x − with HNO 3 represent the most likely route to HNO 3 detection at m/z 62 in our CIMS, other possibilities are the reactions of oxide, superoxide and ozone anions (O x − ) and hydrated O x − with HNO 3 as they have large rate coefficients (> 10 −9 cm 3 molecule −1 s −1 ) and form NO 3 − (Huey, 1996;Wincel et al., 1996;Lengyel et al., 2020): However, when adding O 3 (up to 600 ppbv) to the IMR, we saw no signal that could be attributable to any oxide anion O x − . Figure 7a displays the dependence of the NO 3 − signal at m/z 62 on the O 3 mixing ratio with HNO 3 held constant at 40(±6) ppbv. The dependence of the normalised signal at m/z 62 on [O 3 ] is clearly non-linear. We showed above that the sum of all IO x − anions increases approximately linearly with O 3 mixing ratio while at the same time the behaviour of IO − and IO 3 − is more complex. The sensitivity of HNO 3 detection to changes in O 3 concentration is expected to depend not only on the individual concentrations of each anion in the IMR but also on their respective rate coefficients for reaction with HNO 3 (which are different, see above), and which reflects the expected, approximately exponential dependence of the concentration of IO x − in the IMR on the O 3 concentration. In Eq. (2), [O 3 ] is the O 3 mixing ratio in ppbv and B has a value of 1.515 × 10 −3 per ppbv of O 3 . As shown in Fig. 7b, for a given [HNO 3 ], the parameter A is dependent on the water vapour concentration (i.e. on ratio of signals at m/z 145 and m/z 127, (S 145 /S 127 )) over the range explored and can be parameterised as In these experiments, H 2 O was not added to the thermal dissociation (TD) inlet (this would have increased the retention time of HNO 3 on inlet surfaces, thereby making changes in the m/z 62 difficult to interpret) but directly to the IMR, as described in Sect. 2 and as used during airborne operation of the CIMS. The linear dependence of the signal at m/z 62  Fig. S3.
The positive intercept in Fig. 7b indicates that there is significant sensitivity to HNO 3 detection at m/z 62 in the absence of water in the IMR, implying that IO x − anions can react directly with HNO 3 to form NO 3 − as written in Reactions (R14)-(R16). The increase in the sensitivity to HNO 3 as the water vapour concentration is increased is consistent with the formation of I − (H 2 O) (m/z 145) which reacts more rapidly with O 3 (to form IO 2 − directly) than does I − (Teiwes et al., 2019), thereby increasing the abundance of IO x − in the IMR (see above) and thus the instrument's sensitivity to HNO 3 .
The very strong sensitising effect of ozone and H 2 O vapour can explain why similar instruments to ours observe large signals at m/z 62 when sampling ambient air. Indeed, both O 3 and HNO 3 are ubiquitous and generally present at much higher levels than either NO 3 or N 2 O 5 . Attempts to measure these traces gases using I-CIMS without TD inlets and NO titration (to remove the HNO 3 contribution) will likely result in erroneously high levels of both, especially during the day when lower-tropospheric O 3 and HNO 3 are often at their highest levels. It also explains why laboratory tests (generally carried out without added O 3 or H 2 O) have shown only low (or no) sensitivity to HNO 3 at m/z 62.

Detection of HNO 3 at m/z 190
We now compare the detection of HNO 3 at m/z 62 to its detection at m/z 190, the I − (HNO 3 ) adduct, with various amounts of O 3 present. In Fig. 8 we present the results of an experiment in which a constant flow of HNO 3 (12.5 ppbv) was introduced into the inlet and the ozone mixing ratio was varied from zero to 900 ppbv. We observe a great increase in the signal at m/z 62 as expected (from 9 counts to > 6000 counts). At zero ozone, the signal at m/z 190 is about 1000 counts and is largely background free, making this the preferred mass to monitor HNO 3 in the absence of O 3 . The cross-over point (when the signals at m/z 62 and m/z 190 are equal) is at an ozone mixing ratio of 100 ppbv. At an ozone mixing ratio of 800 ppbv, the signal at m/z 62 is a factor 8.5 larger than that at m/z 190.
Apparent from this figure is the depletion of the signal at m/z 190 as the O 3 mixing ratio increases to values of 800 ppbv, as present, for example, in the lower stratosphere. The solid lines are least-squares fits to the datasets that de- scribe the exponential growth of the m/z 62 signal as O 3 is increased and the exponential decay of m/z 190 over the same range. The reduction in signal at m/z 190 is characterised by an exponential term exp(−0.00046 × [O 3 ]), which means that at 800 ppbv O 3 a ∼ 30 % reduction in sensitivity is observed. The loss of sensitivity at m/z 190 is driven by the loss of I − in the IMR as O 3 is added. At the same time, the depletion of the signal at m/z 127 is weaker, which reflects the fact that only a small fraction of IO x − formed in the IMR are detected, the rest presumably being fragmented in the CDC before being detected as I − .
The fact that some I − is converted to IO x in the IMR at high O 3 levels but is not reflected in the I − signal at m/z 127 has repercussions for normalisation of product ion signals to the primary ion signal whereby the assumption is made that the measured signal at m/z 127 stems only from I − . In our instrument, the loss of I − in the IMR is significant at high levels of O 3 (e.g. 30 % at 800 ppbv O 3 ). For this reason, we normalise the signals using values of the signal at m/z 127 obtained by interpolating between measurements obtained when scrubbing the air. The normalisation problem may occur in other I-CIMS instruments to varying extents, and the degree of bias will depend on the conversion of I − to IO x − and the extent to which IO x − is detected as I − . The potential bias can be circumvented by the continuous addition of a calibration gas that is detected via reaction with I − .

Detection of HNO 3 at m/z 62: the role of acetate anions
We have also evaluated the potential for "unintentional" HNO 3 detection at m/z 62 by its reaction with the acetate anion, CH 3 CO 2 − : The CH 3 CO 2 − anion is the conjugate base of a weak acid (CH 3 C(O)OH) has been utilised to monitor a number of trace gases via proton transfer (Veres et al., 2008). While Veres et al. (2010) generated CH 3 CO 2 − deliberately by passing acetic anhydride through their 210 Po source, in our experiments it is the product (monitored at m/z 59) of the reaction between I − primary ions and either CH 3 C(O)O 2 (from the thermal dissociation of PAN) or CH 3 C(O)OOH. Figure 9a shows the result of a set of experiments demonstrating HNO 3 detection at m/z 62 without (blue data points) and with 3.25 ppbv of CH 3 C(O)OOH (black data points) added to the inlet flow. The initial (non-normalised) signal at m/z 59 from the CH 3 CO 2 − anion in the absence of HNO 3 was 53 500 counts. The presence of 3.25 ppbv CH 3 C(O)OOH (and resultant CH 3 CO 2 − ) results in a ∼ 50fold increase in the sensitivity of the I-CIMS to HNO 3 . We also carried out a few experiments (less systematic) in which PAN (instead of PAA) was added to the IMR and obtained the same results.
Our results disagree with the conclusions of Wang et al. (2014), who saw no increase at m/z 62 when adding PAN to air containing HNO 3 but are consistent with the use of CH 3 CO 2 − as a primary reactant ion to detect HNO 3 at m/z 62 (Veres et al., 2008). Figure 9b indicates that the increase in signal at m/z 62 when adding HNO 3 to a flow of CH 3 C(O)OOH in air is approximately proportional to the reduction in the ion signal at m/z 59. This helps confirm that CH 3 CO 2 − is the ion responsible for the detection of HNO 3 but also indicates that the detection of PAN and CH 3 C(O)OOH via conversion to CH 3 CO 2 − can be compromised when HNO 3 is present in the air sample. Indeed, in many air masses the concentration of HNO 3 can be an order of magnitude greater than that of either PAN or CH 3 C(O)OOH, and given that other abundant trace gases (e.g. organic acids) also react with CH 3 CO 2 − (Veres et al., 2008), further reactions of CH 3 CO 2 − in the ionmolecule reactor regions of I-CIMS instruments may result in a significant bias (to lower values) which would have to be analysed case by case for different instruments. One way to avoid this problem is the continuous addition of isotopically labelled PAN to the inlet (see e.g. Roiger et al., 2011) as the secondary, reactive losses of 12 C and 13 C CH 3 CO 2 − are expected to be identical. Wang et al. (2014) observed that the majority of the m/z 62 signal during the daytime could be removed by addition of NO (0.54 ppmv or 1.3 × 10 13 molecule cm −3 ) to the inlet. At their inlet temperature of 120-180 • C, NO reacts with O 3 with a rate coefficient in the range 6-9 × Figure 9. (a) Detection of HNO 3 at m/z 62 in the presence of 3.25 ppbv PAA (and thus the acetate anion, CH 3 CO 2 ). The blue data points (expanded view in the inset) were obtained in the absence of PAA, whereby detection of HNO 3 at m/z 62 is inefficient. The error bars are 1σ statistical uncertainty in the signal at m/z 62 and 17 % total uncertainty in the HNO 3 mixing ratio. (b) Change in normalised signals at m/z 62 and m/z 59 upon adding HNO 3 for the same dataset (i.e. background corrected signals). The error bars are 1σ statistical uncertainty. 10 −14 cm 3 molecule −1 s −1 , which results in a half-life for O 3 of 500 to 800 ms. Wang et al. (2014) do not mention the residence time of air passing through their heated inlet, but it appears plausible that a substantial fraction of ambient O 3 would have been removed during background measurement, thus decreasing (or removing) sensitivity towards HNO 3 via reactions involving O 3 in the IMR, and leading the authors to conclude that NO 3 was being detected above a lower background than truly present.
To illustrate the potential size of the bias due to HNO 3 when monitoring N 2 O 5 at m/z 62 in field measurements, we take the relative sensitivities (at m/z 62) of our I-CIMS to N 2 O 5 and to HNO 3 in the presence of typical boundary layer mixing ratios of O 3 (50 ppbv) and at typical relative humidity (50 %). Under these conditions, with N 2 O 5 and HNO 3 mixing ratios of 0.2 and 2 ppbv, respectively, we calculate that HNO 3 would account for > 70 % of the signal at m/z 62.

Field measurements
Having shown that HNO 3 is detected by our I-CIMS with reasonable sensitivity when sufficient O 3 is present in ambient air samples, we now examine the signals at m/z 62 obtained in airborne operation of the I-CIMS during two CAFE (Chemistry of the Atmosphere Field Experiment) campaigns of the HALO aircraft. In the CAFE-Africa campaign (2018) the I-CIMS monitored m/z 62 on several flights over the Atlantic west of the African continent. During the 2020 CAFE-EU campaign with HALO over Europe, the I-CIMS additionally monitored m/z 190 (the I − (HNO 3 ) cluster ion), which is selective to HNO 3 . During both campaigns, O 3 and H 2 O (required for analysis of the signal at m/z 62) were routinely measured.

CAFE-Africa
Here we examine the results obtained during a HALO flight as part of the CAFE-Africa mission. The flight in question was the transfer from Sal airport on the Cabo Verde islands (which served as the base station during the mission) back to Germany. During the flight the aircraft flew mainly at high altitudes (13-15 km) so that stratospheric air was sampled at higher latitudes but also made two dives into the free troposphere. The flight track is displayed in Fig. S4. Figure 10 shows a time series of ozone mixing ratios during the flight (panel a) along with the I-CIMS signal at m/z 62 (panel b). In air masses with stratospheric influence (i.e. O 3 values > 100 ppb, 12:20-15:10 UTC) there is an obvious, strong co-variance between these two parameters. However, once corrected for the dependence of the sensitivity of the I-CIMS to O 3 (Eqs. 2 and 3) we obtain the black line representing the mixing ratios of HNO 3 and the covariance is greatly reduced. We also note that, apart from some significant increases at ∼ 11:30 and ∼ 16:00, the HNO 3 mixing ratio decreases slowly throughout the flight, which is the result of HNO 3 generation in the 210 Po source leading to an initially large background signal. The formation of HNO 3 in the 210 Po source has been documented previously (Ji et al., 2020); its level can be reduced by permanently flushing N 2 through the source while keeping the mass spectrometer under operational vacuum. This was not carried out during the CAFE missions on HALO as continuous operation of the instrument (i.e. overnight between flights) was not possible. We note that the background signal at m/z 62 that originates from the polonium source cannot be obtained by scrubbing the air of HNO 3 as this also removes O 3 and thus also sensitivity to HNO 3 at this mass. The signal at m/z 62 (red line) clearly co-varies with O 3 . Following conversion to a mixing ratio (black line) and subtraction of an HNO 3 background signal (dotted line, assuming an exponential decay) originating in the 210 Po source, the solid blue line for HNO 3 is obtained (HNO 3 cor). (c) HNO 3 mixing ratios plotted versus O 3 mixing ratios. The straight black line has a slope of (3 ± 0.5) × 10 −3 and does not take into account the grey data points (O 3 mixing ratio < 100 ppbv).
A roughly exponential decay of the HNO 3 background signal was observed in all of the flights in which m/z 62 was monitored, which presumably reflects depletion of the initially large HNO 3 reservoir which was built up when the I-CIMS was switched off.
A coarse correction of the dataset was thus undertaken by subtracting an exponentially decaying background from the total HNO 3 signal. The resulting HNO 3 mixing ratios are depicted as the blue line in Fig. 10b and plotted against the O 3 mixing ratio in Fig. 10c. Considering only the highaltitude data for which O 3 mixing ratios were > 100 ppbv (stratospheric influence, black data points), we derive a slope of HNO 3 /O 3 = (3 ± 0.5) × 10 −3 (the uncertainty is 2σ , statistical only), which is consistent with previously reported values obtained in airborne measurements of HNO 3 and O 3 in the lower stratosphere (see Popp et al., 2009, and refer-ences therein). We stress that deriving accurate mixing ratios of HNO 3 is not possible with this dataset, and the values obtained are strongly dependent on the background correction. Here, we merely wish to indicate that, while most of the variability in our m/z 62 signal is related to the central role of ozone in the detection scheme (i.e. formation of IO x − ), some covariance between HNO 3 and O 3 remains after correction of the raw data and the slope is roughly in line with that expected. We also do not propose that the correlation of m/z 62 with O 3 proves that the signal can be attributed entirely to HNO 3 . This aspect will be covered in Sect. 4.2. Figure 10b reveals sharp increases in the (backgroundcorrected) HNO 3 mixing ratio when sampling at lower altitudes, noticeably at 11:30-12:00 (3.9 km altitude) and at 15:45-16:10 (4.7 km altitude) and at the end of the flight during descent to Oberpfaffenhofen in Bavaria, Germany. In all cases, these periods of enhanced HNO 3 coincided with higher levels of particles. Back trajectories (HYSPLIT) indicated that, in the 10 d prior to interception by HALO, the air mass sampled at 11:30 had passed over the West African continent (Mauritania, Mali and Niger), whereas the air masses sampled after 16:00 were of European origin. The large, coincidental increase in the HNO 3 mixing ratio and particle mass was a recurrent feature of the CAFE-Africa flights. It is conceivable that the HNO 3 measured by the I-CIMS was a mixture of gas-phase HNO 3 and HNO 3 associated with particles that desorb HNO 3 when passing through the thermal dissociation inlet at 180 • C. This temperature would be sufficient to thermally convert ammonium nitrate to HNO 3 (and NH 3 ) as well as to result in the desorption of HNO 3 that was physisorbed, for example, on chemically aged black-carbon or mineral-dust particles. As we do not know the efficiency with which particles of various diameters enter the TD inlet of the CIMS, we cannot estimate the relative contribution of gas-phase and particulate nitrate to the signal at m/z 62 but indicate that a similar phenomenon may occur in groundbased measurements using TD inlets and may represent an additional source of bias during ambient measurements of NO 3 and/or N 2 O 5 at m/z 62.

CAFE-EU
During the CAFE-EU HALO flights the I-CIMS monitored m/z 190, the I − (HNO 3 ) adduct, as well as NO 3 − at m/z 62. The detection of HNO 3 at m/z 190 varies with the water vapour concentration in the IMR: the response of the HNO 3 signal at m/z 190 to changes in the HNO 3 concentration and in the m/z 145 : m/z 127 ratio (i.e. the relative humidity in the IMR, see Fig. S2) is illustrated in Fig. S5. Figure 11 displays a set of data obtained during a flight on 30 May 2020 on which the HALO aircraft flew a path from southern Germany to the Atlantic (west of Ireland) and back at various altitudes (for flight track see Fig. S6). Figure 11a plots the raw signals measured by the I-CIMS at m/z 62 and m/z 190 as well as the O 3 mixing ratio. Similar to the CAFE- (c) Correlation of the HNO 3 mixing ratios derived from the two masses. The red line is a bivariate fit (slope 1.14 ± 0.05, intercept −0.3 ± 0.3), and the black line is a 1 : 1 line. A large fraction of the HNO 3 measured stems from the polonium source, especially at the beginning of the flight.
Africa dataset, the signal at m/z 62 covaries strongly with the O 3 mixing ratios, which were between ∼ 40 and ∼ 700 ppbv, whereas the raw signals at m/z 62 and m/z 190 (both due to HNO 3 ) bear little resemblance to each other.
Using the calibration parameters described in Sect. 3 and (for m/z 190) in Fig. S3, the signals at m/z 62 and m/z 190 were converted to HNO 3 mixing ratios, depicted in Fig. 11b. Despite the greatly divergent raw signals, the HNO 3 mixing ratios obtained using the different mass-to-charge ratios are in reasonable agreement, both displaying a gradual decrease after take-off at ∼ 08:00 UTC. The high initial level of HNO 3 is largely the result of HNO 3 being formed in the 210 Po source during overnight instrument shut-down (see Sect. 4.1). The HNO 3 mixing ratios observed at m/z 62 and m/z 190 both increase when the aircraft sampled stratospheric air (11:00 to 13:00 and 15:10 to 15:30 UTC). In Fig. 11c HNO 3 mixing ratios derived at m/z 62 and m/z 190 are plotted in a correlation diagram. The slope (1.14 ± 0.05) and intercept (−0.3 ± 0.3) indicate reasonable agreement even when the raw signals are greatly divergent at high levels of O 3 . At the highest levels of O 3 , some differences in the retrieved mixing ratios of HNO 3 using m/z 62 and m/z 190 are observed, which, given the large, O 3 -dependent corrections applied (especially for m/z 62), is not surprising. Our airborne data show that in many (if not most) air masses, m/z 62 provides a measure of HNO 3 rather than NO 3 and N 2 O 5 .

Conclusions
A series of laboratory experiments investigating the origin of signal at m/z 62 when using an I-CIMS has revealed unexpected sensitivity to HNO 3 at this mass-to-charge ratio in the presence of O 3 or peroxyacetic acid (PAA) or PAN. The ozone effect is related to the formation of IO x − , which reacts rapidly with HNO 3 to form NO 3 − thus bypassing the thermodynamic barrier to formation of NO 3 − by direct reaction of HNO 3 with I − . The presence of O 3 at a mixing ratio of 500 ppbv results in a 250-fold increase in sensitivity to HNO 3 at m/z 62. The sensitivity to HNO 3 at this massto-charge ratio was also found to be highly dependent on the concentration of H 2 O in the ion-molecule reactor as this aids formation of IO x − . The sensitivity to HNO 3 at m/z 62 in the presence of PAA is a result of the presence of acetate anions (CH 3 CO 2 − ) as demonstrated previously (Veres et al., 2008). We conclude that measurements of PAN using I-CIMS may be biased to low values if large mixing ratios of HNO 3 (or organic acids) are present and continuous calibration (e.g. with isotopically labelled PAN) is not carried out. Our laboratory experiments indicate that measurements of atmospheric NO 3 and N 2 O 5 at m/z 62 can be heavily biased by the presence of HNO 3 and may explain reports of unexpectedly high daytime mixing ratios of N 2 O 5 . The relative sensitivity at m/z 62 to HNO 3 and N 2 O 5 /NO 3 will vary from one I-CIMS instrument to the next and must thus be analysed case by case.
We have examined signals at m/z 62 during two periods of operation of the I-CIMS on the HALO aircraft: one over the Atlantic west of the African coast and one over Europe. During the flights over Europe HNO 3 mixing ratios derived from m/z 62 NO 3 − and at m/z 190 (I − (HNO 3 )) were in good agreement. The data obtained over the Atlantic indicated that measurements at m/z 62 using a thermal dissociation inlet can be strongly influenced by particulate nitrate that can thermally dissociate (or desorb) to gas-phase HNO 3 .
Data availability. Data measured during the flight campaign CAFE campaigns are available to all scientists agreeing to the CAFE data protocol. The laboratory data underlying the figures are available upon request to the authors.
Author contributions. RD conducted the laboratory experiments, carried out the airborne measurements with assistance from PE and JNC, and analysed the laboratory data with assistance from JNC. The manuscript was written by JNC and RD with contributions from all other authors. JL designed and helped plan the airborne operations.