Missing SO 2 oxidant in the coastal atmosphere? – Evidence from high resolution measurements of OH and atmospheric sulfur compounds

Diurnal and seasonal variations of gaseous sulfuric acid (H 2 SO 4 ) and methane sulfonic acid (MSA) were measured in N.E. Atlantic air at the Mace Head atmospheric research station during the years 2010 and 2011. The measurements utilized selected ion/chemical ionization mass spectrometry (SI/CIMS) with a detection limit for both 5 compounds of 4.3 × 10 4 cm − 3 at 5 min signal integration. The H 2 SO 4 and MSA gas-phase concentrations were analysed in conjunction with the condensational sink for both compounds derived from 3 nm–10 µm (diameter) aerosol size distributions. Accommodation coe ﬃ cients of 1.0 for H 2 SO 4 and 0.12 for MSA were assumed leading to estimated atmospheric lifetimes of the order of 7 min and 25 min, respectively. With 10 the SI/CIMS instrument in OH measurement mode alternating between OH signal and background (non-OH) signal evidence was obtained for the presence of one or more unknown oxidants of SO 2 in addition to OH. Depending on the nature of the oxidant(s) their ambient concentration may be enhanced in the CIMS inlet system by additional production. The apparent unknown SO 2 oxidant was additionally conﬁrmed by direct 15 measurements of SO 2 in conjunction with calculated H 2 SO 4 concentrations. The calculated concentrations were consistently lower than the measured concentrations by a factor 4.8 ± 3.4 when considering the oxidation of SO 2 by OH as the only source of H 2 SO 4 . Both the OH and the background signal were also observed


Introduction
It has been well established that homogeneous oxidation of tropospheric gases is gen-5 erally dominated by reactions with the hydroxyl (OH) radical during daylight hours andin regions with significant nitrogen oxide, NO x , concentrations -with the nitrate (NO 3 ) radical in the absence of sunlight (Stone et al., 2012). Reactions of molecular oxygen, ozone, or peroxy radicals such as HO 2 and RO 2 (R = organic rest molecule) are comparatively slow, with few exceptions, such as NO + HO 2 which recycles OH (e.g., 10 Atkinson et al., 2004). Heterogeneous oxidation (on the surface of aerosol particles and in cloud and fog droplets) is dominated either by reactions with dissolved ozone, hydrogen peroxide, or molecular oxygen, the latter pathway being catalyzed by transition metal ions (Harris et al., 2013;Berresheim and Jaeschke, 1986). However, recent studies have revived an interest in the formation and fate of atmospheric Criegee intermediates (RO 2 species produced from reactions of ozone with alkenes, Calvert et al., 2000) which to this day have eluded direct measurements in the atmosphere since Cox and Penkett (1971) first suggested their potentially important role. Field and laboratory measurements (Stone et al., 2014;Taatjes et al., 2013;Mauldin et al., 2012;Vereecken et al., 2012;Berndt et al., 2012;Welz et al., 2012) as well as theoretical studies (Boy Introduction  Selected ion -chemical ionization mass spectrometry (SI-CIMS) has been pioneered by Eisele and coworkers (Tanner and Eisele, 1995;Tanner, 1993, 1991) for high time resolution measurements of OH, H 2 SO 4 , MSA(g) (gaseous methane sulfonic acid), and other compounds in the troposphere. A large number of field studies both on the ground as well as airborne have been successfully con-5 ducted using this technique and significantly improved our understanding of tropospheric chemistry (e.g., Stone et al., 2012;Huey, 2007;Heard and Pilling, 2003). In some of these studies it has already been conjectured that SI/CIMS may also provide information about the presence of atmospheric oxidants other than OH by analyzing the background signal recordings obtained in the OH measurement mode. Specifically, 10 the identity of those "background X -oxidant(s)" was speculated to be Criegee intermediates because of their observed reactivity towards SO 2 in the measurement system (e.g., Berresheim et al., 2002).
In the present paper we have analyzed 2 yr of SI/CIMS measurements made at Mace Head, Ireland, for significant occurrences of such background signals indicating the 15 presence of one or more unknown oxidants in coastal air which contribute to H 2 SO 4 formation by oxidizing SO 2 (in addition to OH) during day-and nighttime. Furthermore, balance calculations of ambient H 2 SO 4 levels using measured SO 2 , OH, and aerosol particle concentrations have been compared with measured H 2 SO 4 levels. This allowed us to approximate corresponding contributions to ambient H 2 SO 4 levels from 20 oxidation of SO 2 by oxidants other than OH and estimate their relative importance with respect to OH reactivity.

Experimental
A principle scheme of the Mace Head CIMS instrument and its operation is shown in Fig. 1. Similar to previously described systems (Berresheim et al., , 2000 Mauldin et al., 1998). Further details including calibration procedures can be found in Berresheim et al. (2000). Propane (99.95 %, Air Liquide, UK) is introduced into the sample flow through the rear injectors (establishing a mixing ratio of approximately 430 ppmv in the sample flow) 5 to scavenge any OH which might be recycled from peroxy radicals via reaction with nitric oxide, NO. On average, nighttime OH measurements showed no major increase in the background signal compared to the OH signal suggesting any potential interference by trace contaminants in the propane to be negligible. Due to similar rate constants for SO 2 and propane with respect to their reaction with OH (both ca. 1 × 10 −12 cm 3 s −1 at 10 298 K; Atkinson et al., 2004) any (recycled) OH molecules are scavenged by propane instead of SO 2 from this point, i.e., downflow from the rear injectors. Due to the very low NO mixing ratios in marine air at Mace Head   signal was found to tail off to a background level corresponding to the complete removal of OH. Increasing the propane flow did not further alter the BG signal. The total reaction time τ reac,X available to this unknown oxidant "X " to react with SO 2 in the system forming H 2 SO 4 is the time starting when a unit volume of the sample flow passes the position of the first injector pairs until it reaches the end of the atmospheric pressure ionization region, i.e., the 80 µm aperture (see Fig. 1). That time in our system corresponds to 0.45 s, or approximately half a second, which is about six times longer than τ reac,OH . Therefore, the relative importance of X in comparison to the atmospheric oxidation efficiency of OH may have to be downscaled dependent on the properties of X and its potential formation and/or regeneration during the reaction time. This will be examined in detail in Sect. 3.4.
Photolysis frequencies of ozone, J(O 1 D), and of nitrogen dioxide, J(NO 2 ), were measured since September 2010 on top of a 10 m tower next to the laboratory building. Both were exchanged with recalibrated systems on a semiannual basis. Details of the measurement principles and performance of the radiometers have been given by Bohn 15 et al. (2008). SO 2 was measured in May-August 2011 with a Thermo Systems 43i instrument using a heated sample inlet teflon tubing (40 • C) to avoid SO 2 losses due to water condensate. Based on a cycle of 30 min signal and 30 min zero measurements (with an added active charcoal filter) we calculated a 2σ detection limit of 25 pptv for 1 h time integration. Figure 2 shows the mean seasonal cycle of the daily maximum H 2 SO 4 concentration in the marine sector at Mace Head which typically occurred between 10-14 h local time, depending on cloud cover. In general, H 2 SO 4 showed a clear diel variation closely cor-25 related with the OH concentration (Fig. 3 relatively homogeneous mixing ratio of the major precursor, SO 2 , in the marine atmosphere, as shown for a three months period in Fig. 4 (top), and the relatively short lifetime of H 2 SO 4 caused by uptake onto aerosol surfaces. This so called condensational sink (CS) showed also low variability on most days (Fig. 4, bottom). The mean SO 2 mixing ratio in the open ocean sector was 160 (±50) pptv during these summer 5 months. The average atmospheric lifetime of H 2 SO 4 with respect to CS was estimated from SMPS and APS measurements using the approach of Fuchs and Sutugin (1971) and Pandis and Seinfeld (1998) to be on the order of 7 min assuming an accommodation coefficient of 1.0 (Kolb et al., 2010;Hanson, 2005 , and a hygroscopic growth factor of 1.7 (Bialek et al., 2012). Overall, we estimate that CS values can be uncertain by at least a factor of two, mainly due to the uncertainties in the count rates of the SMPS and APS instruments and of the hygroscopic growth factor.

15
At Mace Head we assume that the predominant source for H 2 SO 4 in the marine atmosphere is ultimately biogenic, i.e., the emission and oxidation of dimethyl sulfide (DMS) by OH which yields -via further oxidation of intermediate compounds -the gaseous end products H 2 SO 4 , dimethyl sulfone (CH 3 SO 2 CH 3 , DMSO 2 ), and methane sulfonic acid (CH 3 SO 3 H, MSA) (Berresheim et al., 1995(Berresheim et al., , 1993a. As described in 20 the previous section, the two acid compounds are detectable by SI/CIMS using the same instrumental setting as for the OH measurement. Corresponding seasonal cycles of aerosol MSA and non-sea salt sulfate, nss-SO 4 , have been measured at Mace Head using high-resolution time-of-flight aerosol mass spectrometry (HR-TOF-MS). Both aerosol compounds as well as their concentration ratio show a clear seasonal 25 maximum in summer (Ovadnevaite et al., 2013).
The mean seasonal cycle of peak MSA(g) mixing ratios recorded during the same daily time slot as for H 2 SO 4 and summarized as monthly means is also shown in Fig. 2. Similar to H 2 SO 4 and the aerosol sulfur compounds, the highest gas phase MSA(g) lev-ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?
H. Berresheim et al. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | els in the marine atmosphere were observed during the summer months which corroborates the biogenic origin of H 2 SO 4 measured in this sector. Adopting a sticking coefficient of 0.12 for the aerosol scavenging of MSA(g) (De Bruyn et al., 1994) we obtained an average atmospheric lifetime of approximately half hour (25 min) for this compound.
As for H 2 SO 4 this is somewhat shorter than previously estimated from measurements 5 off the north-western coast of the United States (Berresheim et al., 1993b), however, still within the same order of magnitude. Ammann et al. (2013) have questioned the earlier results obtained by De Bruyn et al. (1994) and Schweitzer et al. (1998) for the MSA(g) accommodation coefficient and suggested preferring a value close to one as reported in the most recent study by Hanson (2005). However, in our view, adopting a unity value would be in contradiction to common observations of a relatively slower decline of atmospheric MSA(g) levels in comparison to H 2 SO 4 in late afternoon and evening hours which has been well documented in previous field studies (e.g., Eisele and Tanner, 1993) and in our present study. Furthermore, as shown already in a previous campaign at Mace Head (Berresheim et al., 2002), ambient MSA(g) levels typically 15 increased with decreasing relative humidity, including at nighttime. Both observations support that the vapor pressure of MSA(g) is significantly higher compared to H 2 SO 4 (e.g., Kreidenweis and Seinfeld, 1988).  Berresheim et al. (2002). They speculated that the missing source might be DMS oxidation with partial production of SO 3 instead of SO 2 as intermediate, which then readily forms H 2 SO 4 with water vapour (Lin and Chameides, 1993). This possibility would also agree with kinetic pathways hypothesized for the DMS + OH oxidation in which CH 3 SO 2 and CH 3 SO 3 are formed 15 as intermediates, both of which decompose thermally to SO 2 and SO 3 , respectively (Berresheim et al., 1995). Alternatively, it was suggested that oxidant(s) in addition to OH might play a significant role in this environment, e.g., stabilized Criegee intermediates (sCI) which recently have been re-evaluated with respect to their potential oxidation of atmospheric SO 2 by Welz et al. (2012) and Mauldin et al. (2012). Studies 20 in an Antarctic coastal location with strong marine DMS emissions Davis et al., 1998) reported similar inconsistencies between measured H 2 SO 4 levels and SO 2 mixing ratios required to close the mass balance based on SO 2 + OH as the only source, even when assuming a very low H 2 SO 4 accommodation coefficient of 0.5. The nitrate radical, NO 3 , is not expected to be of any importance for nighttime 25 SO 2 oxidation in such remote locations including Mace Head, at least not in air from the marine sector .

H 2 SO 4 mass balance and missing SO 2 oxidant in the marine atmosphere
ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?

Electronic structure calculations on halogen oxide reactions with SO 2
Other candidates might be halogen oxide radicals, however, to our knowledge respective rate constants are available in the literature only for the reactions of IO and ClO with SO 2 (e.g., kinetics.nist.gov/kinetics/), which are three and six orders of magnitude smaller compared to k SO 2 +OH , respectively. We have made ab initio transition state en-5 ergy calculations for the reactions of SO 2 with ClO, BrO, IO, and OIO using quantum theory. The hybrid density functional/Hartree-Fock B3LYP method was employed from within the Gaussian 09 suite of programs (Frisch et al., 2009), combined with an appropriate basis set for I (Glukhovtsev et al., 1995) et al., 2000). At this level of theory, the expected uncertainty in the calculated transition 15 state energies should be better than 0.07 eV (Foresman and Frisch, 1996). Spin-orbit effects were ignored since these are present both in the reactant halogen oxide and the transition state. Figure 7   . However, the very large barrier for the OIO + SO 2 reaction (50.1 kJ mol −1 ) means that this reaction is negligibly slow: k(200-400 K) = 5 6.4 × 10 −13 exp(−6400/T ) cm 3 s −1 , and k(293 K) = 2.2 × 10 −22 cm 3 s −1 .
BrO has been observed at a mixing ratio of several parts per trillion during the day at Mace Head . However, the reaction BrO + SO 2 also has a significant barrier (20.4 kJ mol −1 ), and so the reaction is much too slow in the MBL: k(200-400 K) = 5.8 × 10 −14 exp(−2700/T ) cm 3 s −1 , and k(293 K) = 5.6 × 10 −18 cm 3 s −1 .

10
Finally, the TST calculation for ClO + SO 2 , which also has a significant barrier (24.1 kJ mol −1 ), yields k = 5.2 × 10 −14 exp(−3100/T ) cm 3 s −1 . The theoretical rate coefficient at 298 K is therefore 1.5 × 10 −18 cm 3 s −1 , which is in accord with an experimental upper limit of 4 × 10 −18 cm 3 s −1 at this temperature (DeMore et al., 1997). In summary we conclude that none of the halogen oxides considered here react with SO 2 15 fast enough in ambient air to be likely candidates for the missing SO 2 oxidant(s).

Could X be a Criegee radical produced from ozonolysis?
Ignoring the possibility raised by Lin and Chameides (1992) of SO 3 being a major intermediate of DMS + OH oxidation, only the oxidation of SO 2 by sCI remains to be investigated based on current knowledge. If "X " is indeed a Criegee intermediate pro- 20 duced from ozonolysis of alkenes and reacting with SO 2 in the atmosphere and in the CIMS inlet system to produce additional H 2 SO 4 , we can estimate its relative contribution compared to the SO 2 + OH reaction as follows.
As already pointed out in the experimental section we have to account for additional formation of [sCI] cims from alkene + O 3 reactions over the total available residence 25 time of 0.45 s in the atmospheric pressure reaction and ionization region of the CIMS instrument (see Fig. 1). By continuous reaction with SO 2 and ionization of the resulting ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?
H. Berresheim et al. This result is the consequence of the fact that both types of sCI, namely sCI produced in ambient conditions (sCI amb = Prod(sCI) · τ sCI,amb ) and sCI produced inside the CIMS inlet are immediately converted to H Criegee intermediates is estimated to be 0.1 s. As already mentioned, approximately 1 % of the H 2 SO 4 is ionized in the CIMS ionization region. Therefore, the production of sCI in this region indeed yields H 34 2 SO 4 via reaction with 34 SO 2 , of which, however, only 0.5 % is ionized, on average, as this process acts linearly. Consequently, we have to modify Eq. (2) to take into account the reduced ionization probability for H 2 SO 4 pro-20 duced in the ionization region: with t res = 450 ms, t reac = 115 ms, t ion = 335 ms, τ −1 sCI,amb = 1/0.2 s + 4.3 s −1 = 9.3 s −1 . This formalism is identical to that derived for a similar instrument by Berndt et al. (2012) ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?
H. Berresheim et al. (chemical ionization time-of-flight mass spectrometer with atmospheric pressure inlet; CI-APi-TOF-MS). Thus, from Eq. (3) it follows that 34 SO 2 oxidation by sCI contributes a background signal which represents an enhancement of the ambient sCI concentration by a factor EF = 3.6. Therefore, if X is indeed a sCI compound (of the kind considered here), the measurement signal resulting from sCI would have to be 5 weighted by 1 : 3.6 with respect to the OH signal to obtain the corresponding ambient air concentration [sCI]. To compare both compounds with respect to their oxidation efficiency towards SO 2 , the corresponding rate constants must be factored in as well, i.e., k sCI+SO 2 /k OH+SO 2 = 6 × 10 −13 cm 3 s −1 /9 × 10 −13 cm 3 s −1 = 0.67, with k OH+SO 2 (298 K) = 9 × 10 −13 cm 3 s −1 taken from Atkinson et al. (2004) and k sCI+SO 2 adopted for 10 the (monoterpene + SO 2 ) reaction as reported by Mauldin et al. (2012). This means that the oxidation efficiency of those sCI compounds would be only on the order of 1 : 5.4, i.e., 20 % compared to that of OH with respect to SO 2 oxidation, assuming that the CIMS background signal is equal to the OH signal as observed on average in the ambient air measurements at Mace Head (see Fig. 6).

15
These calculations depend strongly on the kinetic parameters for the corresponding sCI reactions. In this work we have adopted rate constants published by Mauldin et al. (2012) and Berndt et al. (2012) for stabilized Criegee intermediates produced from ozonolysis of monoterpenes. However, other studies of smaller Criegee intermediates with low internal energies (CH 2 OO by Stone et al. (2014) and Welz et al. (2012);20 CH 3 CHOO by Taatjes et al., 2013a) suggest much faster reactions of these CI species with both SO 2 and H 2 O, respectively. For a sensitivity test we take the parameters from Taatjes el al. (2013a), (k(CI+SO 2 ) = 6.7 × 10 −11 and 2.4 × 10 −11 cm 3 s −1 , k(CI+H 2 O) = 1 × 10 −14 and an upper limit of 4 × 10 −15 cm 3 s −1 for the anti and syn conformers of CH 3 CHOO, respectively) and neglect the fact, that for the conditions in the 25 CIMS inlet only 80 % of these CI would react with the added 34 SO 2 . We also neglect their unimolecular decomposition whose rate constant is given as an upper limit of 250 s −1 by Taatjes et al. (2013a), since this process would make only a small contribution to our estimates. We find that the oxidation efficiency of such ozonolysis -compared to OH would be approximately 1 : 11.8 for anti and 1 : 13.2 for syn conformers of CH 3 CHOO compared to our earlier estimate of 1 : 5.4, again based on the condition of equal CIMS background and OH signal counts. The relatively small difference between these estimates is a consequence of the fact that both reaction parameters (for CI + SO 2 and CI + H 2 O) are faster in this second estimate. The effect of 5 a faster reaction of CI with SO 2 is almost exactly cancelled out by the faster reactions with H 2 O. For these reasons, if the oxidant(s) X would be such types of stabilized Criegee intermediates, the combined oxidation efficiency of both compounds is estimated to account for a factor of approximately 1.2, increasing the calculated H 2 SO 4 concentra-10 tion based on the SO 2 +OH source alone by only 10-20 %. This is still a major shortfall with respect to the average factor of 4.8 required to match the observed ambient air H 2 SO 4 concentration. Assuming a (rather unlikely) H 2 SO 4 accommodation coefficient as low as 0.5 would reduce this discrepancy by only 30 %. As discussed earlier, we assume that much of the uncertainty remains with the calculation of the condensational 15 sink. However, as yet unknown interferences in the CIMS background signal measurements cannot be entirely ruled out. It appears particularly puzzling that the BG signal frequently tracks the OH signal suggesting that X has similar sources and sinks as OH (Fig. 6). Good candidates for the origin of the CIMS background signal are stabilized Criegee intermediates or iodine oxide (see discussion below). The consequence 20 for the ambient H 2 SO 4 budget is more complex. Either different Criegee intermediates than those studied so far or an entirely different kind of oxidant for SO 2 or a production process converting a sulfur compound other than SO 2 might be still missing in our present account of the H 2 SO 4 concentration in the coastal marine atmosphere.
Recently, Taatjes et al. (2013b) suggested that CH 2 OO might be an important in- 25 termediate in marine air resulting from both ozonolysis of alkenes and photolysis of CH 2 I 2 . Studies by Stone et al. (2014) and Welz et al. (2012)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | events during which both H 2 SO 4 and MSA(g) concentrations increased significantly in conjunction with a major increase in the background signal counts for the X -oxidant(s). A recent successful H 2 SO 4 intercomparison experiment at Mace Head (M. Sipilä and S. Richters, personal communication, 2013) between the CIMS instrument and a CI-APi-TOF-MS instrument has confirmed that the CIMS indeed measures only the con-5 centration of gaseous "free" (monomeric) H 2 SO 4 during nucleation events. With the rapid transition from monomers to multimer clusters in which H 2 SO 4 becomes tied up (confined) and not broken down anymore to the HSO − 4 core ions in the CIMS collision dissociation chamber (Fig. 1) a net decrease in ambient H 2 SO 4 concentrations may therefore be expected. However, as shown in the nucleation event in Fig. 8, (monomer) 10 H 2 SO 4 levels even increased after a certain lag time following the onset of the event. We interpret our observations as strong formation of X -oxidant(s) and OH (perhaps via thermal decomposition of sCI; Berndt et al. (2012), Kroll et al., 2001) followed by rapid oxidation of DMS and SO 2 to form the products H 2 SO 4 and MSA(g). A second, albeit less intensive event was registered on the same day during the evening low 15 tide period ending near sunset (21:30 UTC). Even during this smaller event some increase in OH and X -oxidant levels could be observed. Such coastal nucleation events have previously been shown to be induced by photolysis and photo-oxidation of marine iodine compounds emitted mainly from exposed seaweed during low tide   -Lopez et al., 2006;Alicke et al., 1999). Future studies are required to systematically characterize remaining uncertainties in the CIMS and CS measurements and to verify a possible link between the unknown oxidant(s) and the iodine cycle in the marine atmosphere. 5 We observed a persistent but relatively low H 2 SO 4 background concentration at nighttime (on the order of a few 10 5 cm −3 ) with OH below the detection limit. Also, on some occasions short spikes were observed at nighttime in the background signal during low tide which might suggest short-term emissions of reactive hydrocarbons capable of forming sCI compounds and OH in reactions with ozone. We assume that 10 such processes also happen during daytime but are superimposed by the formation of another major oxidant which shows a similar diurnal pattern like OH. Whether this oxidant might be a Criegee radical with its production mainly determined by strong light-induced emissions of marine hydrocarbon species and/or atmospheric photolysis of iodine species remains an open question. However, we consider it unlikely that α-15 pinene or limonene are present at significant levels in the marine atmosphere. For this reason and also based on the currently available kinetic data for the SO 2 oxidation by sCI compounds resulting from these monoterpenes we conclude that at least those specific sCI radicals are unimportant in comparison with the SO 2 + OH oxidation in the marine atmosphere. In the present work we have shown that the OH background sig- 20 nal measured with the CIMS instrument provides evidence for the presence of one or more unknown oxidants for atmospheric SO 2 in addition to OH. However, as this oxidant X does not significantly react with propane in the CIMS system, the corresponding X-signal must be corrected to account for additional production inside the CIMS inlet system before evaluating its oxidation efficiency towards SO 2 in ambient air. However, ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?

Conclusions
H. Berresheim et al. the proposed oxidation efficiency for SO 2 of stabilized Criegee intermediates from αpinene or limonene in forested environments as well. ACPD 14,2014 Missing SO 2 oxidant in the coastal atmosphere?