Seasonal variations of triple oxygen isotopic compositions of atmospheric sulfate, nitrate, and ozone at Dumont d'Urville, coastal Antarctica

Abstract. Triple oxygen isotopic compositions (Δ17O  = δ17O − 0.52  ×  δ18O) of atmospheric sulfate (SO42−) and nitrate (NO3−) in the atmosphere reflect the relative contribution of oxidation pathways involved in their formation processes, which potentially provides information to reveal missing reactions in atmospheric chemistry models. However, there remain many theoretical assumptions for the controlling factors of Δ17O(SO42−) and Δ17O(NO3−) values in those model estimations. To test one of those assumption that Δ17O values of ozone (O3) have a flat value and do not influence the seasonality of Δ17O(SO42−) and Δ17O(NO3−) values, we performed the first simultaneous measurement of Δ17O values of atmospheric sulfate, nitrate, and ozone collected at Dumont d'Urville (DDU) Station (66°40′ S, 140°01′ E) throughout 2011. Δ17O values of sulfate and nitrate exhibited seasonal variation characterized by minima in the austral summer and maxima in winter, within the ranges of 0.9–3.4 and 23.0–41.9 ‰, respectively. In contrast, Δ17O values of ozone showed no significant seasonal variation, with values of 26 ± 1 ‰ throughout the year. These contrasting seasonal trends suggest that seasonality in Δ17O(SO42−) and Δ17O(NO3−) values is not the result of changes in Δ17O(O3), but of the changes in oxidation chemistry. The trends with summer minima and winter maxima for Δ17O(SO42−) and Δ17O(NO3−) values are caused by sunlight-driven changes in the relative contribution of O3 oxidation to the oxidation by HOx, ROx, and H2O2. In addition to that general trend, by comparing Δ17O(SO42−) and Δ17O(NO3−) values to ozone mixing ratios, we found that Δ17O(SO42−) values observed in spring (September to November) were lower than in fall (March to May), while there was no significant spring and fall difference in Δ17O(NO3−) values. The relatively lower sensitivity of Δ17O(SO42−) values to the ozone mixing ratio in spring compared to fall is possibly explained by (i) the increased contribution of SO2 oxidations by OH and H2O2 caused by NOx emission from snowpack and/or (ii) SO2 oxidation by hypohalous acids (HOX  =  HOCl + HOBr) in the aqueous phase.


Introduction
Triple oxygen isotopic compositions ( 17 O = δ 17 O − 0.52 × δ 18 O) of atmospheric sulfate and nitrate have shown the potential to probe the relative importance of various oxidation pathways involved in their formations (e.g., Michalski et al., 2003;Lee and Thiemens, 2001).Atmospheric ozone (O 3 ) possesses high 17 O values of approximately 26 ‰ (Krankowsky et al., 1995; Johnston and  S. Ishino et al.: 17 O values of atmospheric sulfate and nitrate in coastal Antarctica Thiemens, 1997;Vicars et al., 2012;Vicars and Savarino, 2014), in contrast to most of the oxygen bearing compounds such as O 2 and H 2 O which have 17 O values of approximately 0 ‰ (Barkan andLuz, 2003, 2005).The oxygen atoms of O 3 are directly or indirectly transferred to sulfate and nitrate through the various oxidation pathways of their precursors, SO 2 and NO x (= NO + NO 2 ) respectively.Therefore, O 3 oxidation produces sulfate and nitrate possessing high 17 O values, whereas oxidation by OH, RO 2 , and H 2 O 2 produces sulfate and nitrate with low 17 O values.In general, the 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values can be prognosticated using the given 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values for each reaction pathway and the estimates of relative contribution of each oxidation pathways by atmospheric chemistry transport models (e.g., Alexander et al., 2009;Morin et al., 2011;Sofen et al., 2011).Considering the higher HO x levels in summer due to enhanced photochemical activity relative to winter, 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values are expected to show the specific seasonal trend with summer minima and winter maxima in the mid-to high-latitude regions.By comparing those expected values to observations, the missing processes for sulfate and nitrate formation in the models have been proposed.For example, McCabe et al. (2006) performed a year-round observation of 17 O(SO 2− 4 ) values at Alert, Canada, showing lower 17 O(SO 2− 4 ) values during winter compared to the calculated values using a model from Feitcher et al. (1996); McCabe et al. (2006) suggested a 2-fold overestimate of O 3 oxidation in sulfate formation during winter in the Northern Hemisphere and pointed out a 10-18 % contribution of metal-catalyzed O 2 oxidation in aqueous phase.An observation of 17 O(NO − 3 ) values in Alert by Morin et al. (2008) revealed significantly higher values during spring compared to the calculated values, which is expected to be a result of NO 2 oxidation by BrO.Savarino et al. (2013) also had shown significantly higher 17 O(NO − 3 ) values at the marine boundary layer compared to an estimate of the chemical transport model by Alexander et al. (2009), while the mismatch was fixed by considering a contribution of BrO in nitrate formation.The 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values are thus used to examine the lack of oxidation schemes in the models.However, the observational data of 17 O(SO 2− 4 ) and 17 O(NO − 3 ) to constrain these models are still not enough to demonstrate the theoretical assumptions of the controlling factors of 17 O signatures in these model estimations (e.g., Morin et al., 2011).
One of the important parameters which have to be confirmed by observations is 17 O value of ozone, because of the possibility that the 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values are influenced not only by changes in oxidation chemistry but also variations in 17  3 ) values (Alexander et al., 2009;Morin et al., 2011;Sofen et al., 2011).Those varied 17 O(O 3 ) values were assumed based on the observations using the cryogenic collection method (Johnston and Thiemens, 1997;Krankowsky et al., 1995), which showed highly varied values of 6-54 ‰.However, the recent observations of 17 O(O 3 ) values using the nitrite-coated filter method -which was developed by Vicars et al. (2012) and applied at Grenoble, the R/V Polarstern Campaign (Vicars and Savarino, 2014), and Dome C (Savarino et al., 2016)  3 ) values at coastal and inland Antarctica, Dumont d'Urville (DDU) and Dome C, showing an austral summer minimum and a winter maximum (Erbland et al., 2013;Frey et al., 2009;Savarino et al., 2007).This trend is mainly explained by increased NO x oxidation by OH and RO 2 under solar radiation in summer relative to winter when NO x oxidation is dominated by the reaction transferring the oxygen atoms from O 3 to nitrate in dark polar winter, when assuming the values are mainly controlled by oxidation chemistry.The similar trend following the solar cycle is expected for 17 O(SO 2− 4 ) values, which is mainly controlled by the relative importance of SO 2 oxidation by OH, H 2 O 2 , and O 3 .The first attempt of the monthly scale observation of 17 O(SO 2− 4 ) values at Dome C showed the increasing trend from January to June (summer to winter) and the decreasing trend from October to next January (spring to summer) despite the unexpected decline during midwinter, July-August (Hill-Falkenthal et al., 2013).Nevertheless, the variability of 17 O(SO 2− 4 ) values throughout the year is larger than that which has ever been observed on earth (Lee and Thiemens, 2001;Li et al., 2013;McCabe et al., 2006) 3 ) values as well as to provide a set of data that can be used to constrain the chemistry transport model scheme in the future study.The series of oxidant observations such as O 3 (Legrand et al., 2009(Legrand et al., , 2016a)), HO x (Kukui et al., 2012(Kukui et al., , 2014) ) and NO x (Grilli et al., 2013) have demonstrated that due to the katabatic winds, air masses on the East Antarctic plateau enriched in oxidants produced via reactive nitrogen emission from surface snow and subsequent interactive reactions between HO x , NO x , and RO 2 , are frequently exported to DDU.Legrand et al. (2009Legrand et al. ( , 2016a) ) also suggested that the influence of bromine chemistry is much less significant in East Antarctica compared to a western coastal site such as Halley and Neumayer.However, besides the framework of that project, model-aided analyses of 17 O(SO 2− 4 ) observations by Chen et al. (2016) pointed out the significant contribution (33-50 %) of SO 2 oxidation by hypohalous acids (HOX = HOCl and HOBr) on total sulfate production in the remote marine boundary layer including the Southern Ocean, indicating the possibility of its influence in the coastal Antarctic site.Hence, we then aimed to reveal any features of the oxidation mechanisms of their precursors based on the seasonality of 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values at this specific site.
2 Samples and analytical methods 2.1 Sampling site and aerosol sample collection 2.1.1Sampling site Samples were collected at DDU (66 • 40 S, 140 • 01 E; 40 m a.s.l.), located on a small island 1 km off the coast of Antarctica.The climate of DDU is described in Konig-Langlo et al. (1998).Compared with other parts of Antarctica, DDU is temperate, with temperatures ranging from −30 to 5 • C throughout the year.Most parts of the island are free of snow, and the sea ice disappears completely during most of the summers.Recent observations of surface ozone and the OH radical (Legrand et al., 2016a;Kukui et al., 2012) revealed that at DDU, O 3 and OH levels are approximately 2 and 10 times higher, respectively, compared to the Palmer Station (Jefferson et al., 1998), due to the transportation of air masses influenced by the snowpack emission of reactive nitrogen species on the East Antarctic plateau and the subsequent oxidant productions.

Aerosol sample collection
Aerosol samples were collected using a high-volume air sampler (HVAS; General Metal Works GL 2000H Hi Vol TSP; Tisch Environmental, Cleves, OH, USA).Coarse (> 1 µm) and fine (< 1 µm) particles were collected separately using a four-stage cascade impactor and a backup glass fiber filter, respectively.The slotted 12.7 cm × 17.8 cm glass fiber filters were mounted on the cascade impactor, while the 20.3 cm × 25.4 cm glass fiber filters were used for backup.The HVAS was placed on a platform 1 m above ground, 50 m from the coast, and 20 m away from the closest building.Aerosol collection was carried out at weekly intervals, with flow rates of ∼ 1.5 m 3 min −1 , yielding an average pumped air volume of 15 000 m 3 per sample.Samples collected between January 2011 and January 2012 were used for this study.Once per month, a field blank was checked by mounting filters onto the filter holder and running the cascade impactor for 1 min.
After each collection period, the filters were removed from the cascade impactor inside a clean chemical hood; they were wrapped in aluminum foil and stored in plastic bags at −20 • C. The four filters on the impactor stage were grouped together as coarse particle samples, while the backup filters were kept as fine particle samples.Samples were transported back to the Université Grenoble Alpes (France) for chemical and isotopic analyses, while frozen.

Quantification of ionic species
The soluble compounds in the aerosols were extracted with ultrapure water (Millipore filter, 18 M cm; EMD Millipore, MA, USA) according to the process described in Savarino et al. (2007); more than 98 % of the initial water volume was recovered.Field blank filters were processed in the same way.
Small aliquots of these sample solutions were taken for quantification of ionic species.Anion (Cl − , NO − 3 , SO 2− 4 ) and sodium (Na + ) concentrations were analyzed using the ion chromatography systems described in Savarino et al. (2007) and Jourdain and Legrand (2002), respectively.
Atmospheric concentrations of these ionic species were calculated using the aerosol loading for each filter, the mean filter blank values, and the air volume pumped through the filter.The air volume was corrected to standard temperature and pressure (T = 273.15K, p = 101 325 Pa) based on the meteorological data from DDU provided by Meteo France.The uncertainties for atmospheric concentrations were calculated by propagating the typical uncertainty of the ion chromatography analysis (5 %) and the standard deviation (1σ ) of filter blank values.

Oxygen isotopic analyses of sulfate and nitrate in aerosols 2.2.1 Definition of triple oxygen isotopic compositions
Given the two isotope ratios, notated as 17 R (= 17 O / 16 O) and 18 R (= 18 O / 16 O), stable oxygen isotope ratios are conventionally scaled using a delta (δ) notation as follows.
where R VSMOW denotes the isotope ratio of the standard material, Vienna Standard Mean Ocean Water (VSMOW), and x is 17 or 18.Despite the robust relationship of the massdependent law (δ 17 O = 0.52 × δ 18 O) in most of the oxygencontaining species (e.g., O 2 and H 2 O), atmospheric ozone does not follow mass-dependent fractionation and possesses a significant positive 17 O (= δ 17 O − 0.52 × δ 18 O) inherited from mass-independent fractionation associated with its formation process (Gao and Marcus, 2001).Since non-zero 17 O values can be observed in various atmospheric species bearing oxygen atoms inherited from O 3 (e.g., sulfate and nitrate), the 17 O signature is a powerful tracer used to investigate the relative contribution of O 3 to oxidation processes.

Oxygen isotopic analysis of sulfate and data correction
All 17 O values of sulfate were measured with an isotope ratio mass spectrometer (IRMS) (MAT253; Thermo Fisher Scientific, Bremen, Germany), coupled with an inhouse measurement system at the Tokyo Institute of Technology.The measurement system for 17 O(SO 2− 4 ) follows Savarino et al. (2001) with modifications described in several studies (Schauer et al., 2012;Geng et al., 2013).Briefly, 1 µmol of sulfate was separated from other ions using an ion chromatography system (Dionex ICS-2100, Thermo Fisher Scientific).The sample was pumped at a flow rate of 1 mL min −1 to the pre-concentration column (Dionex Ion-Pac AG15, 4 mm × 50 mm, Thermo Fisher Scientific) to collect all anions contained in the sample.After the sample was loaded on the pre-concentration column, the anions were eluted by KOH eluent to the guard column (Dionex Ion-Pac AG19, 4 mm × 50 mm, Thermo Fisher Scientific) and the separation column (Dionex IonPac AS19, 4 mm × 250 mm, Thermo Fishier Scientific).The separated sulfate was collected in a vial using a fraction collector (CHF122SC, AD-VANTEC).The sulfate was chemically converted to silver sulfate (Ag 2 SO 4 ).This Ag 2 SO 4 powder is transported in a custom-made quartz cup, which is dropped into a furnace at 1000 • C within a high temperature conversion elemental analyzer (TC/EA; Thermo Fisher Scientific, Bremen, Germany) and thermally decomposed into O 2 and SO 2 .Gas products from this sample pyrolysis are carried by ultrahighpurity He (> 99.99995 % purity; Japan Air Gases Co., Tokyo, Japan), which is first purified using a molecular sieve (5Å) held at −196 • C (Hattori et al., 2015).The gas products O 2 and SO 2 are carried through a cleanup trap (trap 1) held at −196 • C to trap SO 2 and trace SO 3 , while O 2 continues to another molecular sieve (5Å) in a 1.59 mm o. d. tubing trap (trap 2) held at −196 • C to trap O 2 separately from the other gas products.The O 2 is purified using a gas chromatograph with a CP-Molsieve (5Å) column (0.32 mm i. d., 30 m length, 10 µm film; Agilent Technologies Inc., Santa Clara, CA, USA) held at 40 • C before being introduced to the IRMS system to measure m/z = 32, 33, and 34.The interlaboratory calibrated standards (Sulf-α, β, and ε; Schauer et al., 2012) were used to assess the accuracy of our measure- ments; our values were in good agreement with published ones (Fig. 1).As discussed by Schauer et al. (2012), this method results in the oxygen isotope exchange between the O 2 products and the quartz cups as well as the quartz reactor, which shifts δ 17 O, δ 18 O, and thus 17 O measurements.The shift in the 17 O(SO 2− 4 ) value is corrected by estimating the magnitude of the oxygen isotope exchange with quartz materials, whose 17 O value is assumed to be approximately 0 ‰ (Matsuhisa et al., 1978).The intercept of −0.03 in Fig. 1  Since sea salt sulfate aerosols (ss-SO 2− 4 ) are of little importance to the atmospheric sulfur oxidation processes (i.e., 17 O(ss-SO 2− 4 ) = 0 ‰), both total sulfate concentrations and 17 O values were corrected for their ss-SO 2− 4 component to obtain their non-sea salt sulfate (nss-SO 2− 4 ) content using Eqs.( 2) and ( 3 where "total" is the quantity measured by ion chromatography, corresponding to the sum of ss-and nss-SO 2− 4 components, and k is the mass ratio of [SO 2− 4 ] / [Na + ] in sea water (0.25;Holland et al., 1986).To take into account the sea salt chemical fractionation processes that affect the Antarctic region in winter, when temperatures drop below −8 • C in the presence of sea ice (Wagenbach et al., 1998), we used a k value of 0.13 ± 0.04, estimated previously by Jourdain and Legrand (2002) from the winter average at DDU and confirmed by our own dataset; this was applied to samples collected from May to October.Note that the sea salt fractionation is a chemical fractionation and is different from an isotopic fractionation.Equation ( 3) is the isotope mass balance equation between ss-and nss-SO 2− 4 , with 17 O(ss-SO 2− 4 ) = 0 ‰.The total uncertainties for 17 O(nss-SO 2− 4 ) values were calculated using the precision of 17 O measurement and the uncertainty of k value, resulting in the uncertainty of ±1.9 ‰ at maximum.The propagated error for both [nss-SO 2− 4 ] and 17 O(nss-SO 2− 4 ) values are shown in the Supplement.
The measurement of 17 O(SO 2− 4 ) values were only performed for the fine mode samples because the sulfate in the coarse mode samples consists of more than 80 % ss-SO 2− 4 .The influence of ornithogenic soil emissions on Na + and SO 2− 4 concentrations was not taken into account since it mainly affects supermicron (coarse) aerosols (Jourdain and Legrand, 2002).

Oxygen isotopic analysis of nitrate
The 17 O value of nitrate was measured simultaneously with δ 18 O and δ 15 N values using a bacterial denitrifier method (Casciotti et al., 2002) coupled with an IRMS measurement using our in-house peripheral system at the Université Grenoble Alpes (Morin et al., 2009).All nitrates in our samples were converted to N 2 O via bacterial denitrification.This N 2 O was introduced to the measurement system, separated from CO 2 , H 2 O, and other volatile organic compounds, and pre-concentrated in a cold trap.The trapped N 2 O was converted into O 2 and N 2 by pyrolysis at 900 • C, using a gold tube furnace, followed by separation of O 2 and N 2 via a 10 m Molsieve (5Å) gas chromatography column before being introduced to the IRMS system.Measurements were performed simultaneously for samples equivalent to 100 nmol nitrate as well as a subset of international nitrate reference materials (US Geological Survey 32, 34, and 35, as well as their mixtures) for correction and calibration of 17 O and δ 18 O values relative to VSMOW and δ 15 N values relative to air N 2 .Analytical uncertainty was estimated based on the standard deviation of the residuals from a linear regression between the measured reference materials and their expected values.The uncertainties (1σ ) for 17 O(NO − 3 ) and δ 15 N(NO − 3 ) were 0.4 and 0.3 ‰, respectively.

Sampling and analytical methods of the oxygen isotopic composition of ozone
The sampling and isotopic analysis of surface ozone were performed by coupling the nitrite-coated filter method with the nitrate isotopic measurements described in Vicars et al. (2012Vicars et al. ( , 2014)).The principle of ozone collection underlying this technique is the filter-based chemical trapping of ozone via its reaction with nitrite: During Reaction (R1), one of the three oxygen atoms of nitrate is transferred from one of the two terminal oxygen atoms of ozone, while the other two oxygen atoms are derived from the reagent nitrite.Since the 17 O signature of ozone is only located on the terminal atoms of ozone (Bhattacharya et al., 2008;Janssen and Tuzson, 2006), simple mass balance implies that 17 O(O 3 ) term is 2/3 of 17 O(O 3 ) bulk .Thus, 17 O(O 3 ) term values can be inferred using the simple mass balance of Eq. ( 4): where 17 O(NaNO 2 ) of the reagent is confirmed to be zero (Vicars et al., 2012).Therefore, the 17 O value of ozone can be determined from the oxygen isotopic composition of nitrate produced on the coated filter via Reaction (R1), determined by the same measurement system described above.
Ozone sampling was carried out by pumping ambient air, using a low-volume vacuum pump (Model 2522C-02; Welch, IL, USA), through a glass fiber filter (∅ 47 mm, GF/A type; Whatman, UK) pre-coated with a mixture of NaNO 2 , K 2 CO 3 , and glycerol.Sampling was conducted once per week from May 2011 to April 2012, with 24-48 h sampling intervals.After sampling, filter samples and procedural blanks were extracted in 18M water.Any unreacted nitrite reagent was removed using the reaction with sulfamic acid and neutralized later with NaOH solutions (Granger and Sigman, 2009;Vicars et al., 2012).The sample solutions were stored in the dark at −20 • C and transported back to Grenoble.After the nitrate concentration analysis using a colorimetric technique (Frey et al., 2009), the isotopic analysis of nitrate (i.e., ozone) was performed using the same protocol as the nitrate isotope analysis.In addition to the isotope measurements of ozone, we aligned the mixing ratio of surface ozone to the weekly average using data reported in Legrand et al. (2016a) to fit the time resolution of our aerosol sampling.

Complementary analyses
To investigate relationships between the origins of the air masses and the 17 O signatures of sulfate and nitrate, transport pathways of sampled air masses were analyzed using the NOAA's HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model (Stein et al., 2015).The model was used with NCEP-NCAR reanalysis data fields using a regular 2.5 • × 2.5 • longitude-latitude grid.Five-day backward trajectories for air masses arriving at the DDU at an altitude www.atmos-chem-phys.net/17/3713/2017/Atmos.Chem.Phys., 17, 3713-3727, 2017 of 40 m a.s.l. were computed twice per day for each day during sampling periods.
The sea ice area fraction around the Antarctic continent was derived from the Advanced Microwave Scanning Radiometer on-board NASA's Earth Observing System Aqua satellite using the ARTIST sea ice algorithm (Kaleschke et al., 2001).The contact times of these air masses with the Antarctic continent and sea ice were calculated using fiveday backward trajectories and sea ice area fractions.

Sulfate
Seasonal variations in atmospheric concentrations and 17 O values of SO 2− 4 are shown in Fig. 2a.Atmospheric concentrations of nss-SO 2− 4 showed a clear seasonal trend.The [nss-SO 2− 4 ] had a maximum of up to ∼ 280 ng m −3 from January to February, corresponding to the austral summer period, but decreased to a background level (∼ 10 ng m −3 ) during the May to August winter period before increasing as summer returned.This trend in [nss-SO 2− 4 ] at coastal Antarctic sites results from enhanced marine biogenic activity emitting dimethyl sulfide (DMS) in circum-Antarctic regions in summer, as has often been reported previously (e.g., Wagenbach et al., 1998;Minikin et al., 1998;Jourdain and Legrand, 2002;Preunkert et al., 2008).As Antarctica is surrounded by ocean, DMS is the major source of atmospheric non-sea salt sulfur (Minikin et al., 1998;Jourdain and Legrand, 2002).Interestingly, a sample from 18-25 July had an anomalously high value of 46 ng m −3 , four times the monthly mean level for July (∼ 12 ng m −3 ).
The 17 O(nss-SO 2− 4 ) values showed the reverse trend, with a summer minimum and a winter maximum.The 17 O(nss-SO 2− 4 ) value increased from 1.0 ‰ observed in January to a maximum of 3.4 ‰ at the end of June, decreasing to 0.9 ‰ in December.The annual weighted mean value of 17 O(nss-SO 2− 4 ) was 1.4 ± 0.1 ‰.Higher values (greater than 2 ‰) were generally observed during April to July, but the anomalous peak from 18-25 July was characterized by a low 17 O(nss-SO 2− 4 ) value of 0.9 ‰.Consequently, the monthly mean value had a maximum in July (2.6 ± 0.6 ‰) when the 18-25 July data were excluded.

Nitrate
Seasonal variations in atmospheric concentrations and 17 O values for nitrate are shown in Fig. 2b.Nitrate concentrations increased to 55 ng m −3 in January but gradually decreased to less than 10 ng m −3 in March to May.In July, a significant peak of 28 ng m −3 was observed, followed by a seasonal increase as summer returned.The 17 O(NO − 3 ) values showed a simple seasonal variation with a summer minimum and a winter maximum. 17O(NO − 3 ) increased from 27 ‰ in Jan- 17 O values of ozone include samples collected from January to April 2012.uary to over 40 ‰ in July, decreasing moderately to a minimum value of 23 ‰ in December.These trends in nitrate concentrations and 17 O(NO − 3 ) values are consistent with those observed at this site 10 years ago (Savarino et al., 2007).

Ozone
Daily averaged ozone mixing ratios are presented in Fig. 2c; these exhibit a distinct seasonal variation with a summer minimum and a winter maximum.The minimum ozone mixing ratio was observed in January, having a value lower than 10 ppbv, while the maximum was observed during July to August, having a value higher than 35 ppbv.From November to December, sudden increases in ozone levels to values over 30 ppbv were observed a few times, consistent with seasonal trends for ozone at DDU (Legrand et al., 2009) 1.

17 O values of sulfate
Since SO 2 quickly exchanges its oxygen atoms with abundant water vapor in the atmosphere, the 17 O(SO 2 ) value is assumed to be 0 ‰ (Holt et al., 1983).Thus, the 17 O(SO 2− 4 ) value is only dependent on the oxidation pathway of SO 2 to SO 2− 4 .SO 2 oxidation by OH ( 17 O(OH) = ∼ 0 ‰) in the gas phase produces sulfuric acid (H 2 SO 4 ) which possesses the 17 O(SO 2− 4 ) value of approximately 0 ‰.
SO 2 can also dissolve into the aqueous phase on aerosol surfaces, where it can be oxidized by O 3 , H 2 O 2 , or metalcatalyzed oxidation by O 2 to form sulfate (Seinfeld and Pandis, 2006).Given that 17 O(O 3 ) bulk values of approximately 26 ‰ have been observed, the 17 O(SO 2− 4 ) value of sulfate produced by ozone should be around 6.5 ‰, based on a 17 O signature transfer factor of 0.25 (Savarino et al., 2000).
Additionally, aqueous phase SO 2 oxidation by hypohalous acids (HOX = HOCl and HOBr) has been proposed as one of the major reactions in the marine boundary layer (Vogt et al., 1996;von Glasow et al., 2002).Details are discussed in Sect.4.3.Thus, 17 O(SO 2− 4 ) of nss-SO 2− 4 results from a subtle balance between various oxidation reactions, with each one transferring a specific amount of 17 O signature to sulfate.

17 O values of nitrate
The 17 O(NO − 3 ) value is dependent on both the 17 O(NO 2 ) value and the oxidation pathways of NO 2 to NO − 3 .The 17 O(NO 2 ) value is determined by the relative contribution of NO oxidation pathways during the following photochemical cycle.
Since all non-zero 17 O of ozone are positioned in the terminal oxygen atoms (Bhattacharya et al., 2008), which preferentially react with NO (Savarino et al., 2008), the NO 2 formed by ozone exhibits a higher isotopic value than the bulk 17 O(O 3 ).On the other hand, the NO + RO 2 reaction produces nitrate with a lower 17 O(NO − 3 ) value because 17 O(RO 2 ) is approximately 0 ‰ (Morin et al., 2007).NO 2 is then converted into nitrate through one of the following reactions.
It has been pointed out that BrO plays a significant role in both NO and NO 2 oxidation in the marine boundary layer (Savarino et al., 2013) through the following reactions: The oxidation by BrO may also produce NO 2 and nitrate with high 17 O values (Morin et al., 2007), because BrO is thought to possess the terminal oxygen atom of ozone (Bhattacharya et al., 2008).Note that these reactions are thought to have little importance (2 % at maximum) on the Antarctic plateau during the austral summer due to low BrO levels of up to 2-3 pmol mol −1 (Frey et al., 2015;Savarino et al., 2016).Following the principle that two of the three oxygen atoms in NO − 3 come from NO 2 and one arises through conversion  Barkan and Luz (2005).
of NO 2 to NO − 3 , the 17 O(NO − 3 ) value of nitrate produced by each pathway can be expressed as Eq. ( 5).

O NO
For a given value of 17 O(NO 2 ), the NO 2 + OH pathway produces the lowest 17 O(NO − 3 ) value, while the NO 3 + RH pathway or BrONO 2 hydrolysis produce the highest 17 O(NO − 3 ) values.

General trend of seasonal variations in 17 O values of sulfate and nitrate
The 17 O signatures of atmospheric sulfate and nitrate originate from the oxygen transfers from ozone via oxidation of their precursors.Thus, changes in 17 O(O 3 ) values likely affect both 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values.However, given the small seasonal variability of the 17 O(O 3 ) bulk (ca. 5 ‰) and assuming that all oxygen atoms transferred to NO x are from the terminal oxygen of ozone, the expected variability of 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) should not exceed 1.3 and 7.5 ‰, respectively.Clearly, these upper limits do not explain the 2.5 and 19 ‰ seasonal variability observed in 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) at DDU, respectively.Furthermore, the seasonal variation in 17 O(O 3 ) values, with a summer maximum and a winter minimum, is the reverse pattern compared to 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values, with a summer minima and a winter maxima.These inconsistencies suggest that variability in 17 O(O 3 ) values is not the major factor influencing the seasonal variations in 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values.
Meanwhile, 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values are dependent on the relative importance of various oxidation pathways involved in their formation as described in the pre-vious sections.Since the relative importance of these oxidation pathways is sensitive to the relative concentrations of oxidants in the atmosphere and there is seasonal variation for ozone mixing ratios at a continental scale (Crawford et al., 2001;Legrand et al., 2009), the mixing ratio of ozone is expected to correlate with 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values.Indeed, 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values, as well as ozone mixing ratios, all display similar seasonal variations, as shown in Fig. 2. The seasonal variation in the ozone mixing ratios at DDU is generally explained by the accumulation of ozone in winter and its photochemical destruction in summer (Legrand et al., 2009(Legrand et al., , 2016a)), which induces the production of HO x , RO x , and H 2 O 2 in the summer period.Therefore, we propose that seasonal variations in 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) result from a shift in oxidation pathways from O 3 to HO x , RO x , and H 2 O 2 .Decreases in 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values are caused by the combined effects of the decrease in the ozone concentration and the decrease in the transfer efficiency of 17 O(O 3 ) to the final products.Thus, the changes in relative concentrations of O 3 vs.HO x , RO x , and H 2 O 2 , along with the changes in sunlight level, are the main factors controlling the seasonal variations of 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values.A similar seasonal variation in 17 O(nss-SO 2− 4 ) values has been observed at Dome C, an inland Antarctic site (Hill-Falkenthal et al., 2013).However, the 17 O(nss-SO 2− 4 ) values observed at Dome C significantly declined in July and August, in contrast to our observations that showed only a single significant decline in 17 O(nss-SO 2− 4 ) values during the period of 18-25 July.This low 17 O(nss-SO 2− 4 ) sample is also characterized by high nss-SO 2− 4 concentration (Fig. 2a).We don't have any evidence of contamination from station activities or laboratory works.The preliminary result of the sulfur isotope analysis of sulfate in the same sample, showing a δ 34 S value of 17.6 ‰ (to be published), suggests to 17 O(nss-SO 2− 4 ) values, the slopes between 17 O(NO − 3 ) values and the ozone mixing ratios for an intercept fixed to 18.2 show of a less difference between spring and fall, with values of 0.54 ± 0.02 and 0.57 ± 0.01, respectively.However, 17 O(nss-SO 2− 4 ) is also sensitive to HO x , RO x , and H 2 O 2 .If the relative abundance of O 3 vs.HO x , RO x , and H 2 O 2 are only regulated by changes in solar irradiation as discussed in Sect.4.2, then the slopes between the 17 O values for sulfate and nitrate against ozone mixing ratios should be the same for spring and fall.The different slopes observed for sulfate in spring and fall indicate the effects of various oxidation processes, with decreasing 17 O(nss-SO 2− 4 ) in spring and/or increasing 17 O(nss-SO 2− 4 ) in fall.There are several processes that could explain such a spring/fall difference.
One possible explanation involves the influence of NO x emissions from snowpack covering the East Antarctic plateau (Davis et al., 2001;Crawford et al., 2001;Chen et al., 2001).The Antarctic atmosphere is strongly affected by NO x emissions from snowpack, starting at the beginning of spring with the return of sunlight.These snow NO x emissions subsequently enhance both O 3 and OH productions, with OH being in greater proportion than O 3 .This is par-ticularly true at DDU in summer, where katabatic air from the East Antarctic plateau causes the ozone mixing ratio to be in the range of 10-40 ppbv (Legrand et al., 2009(Legrand et al., , 2016a)), with a mean OH concentration of 2.1 × 10 6 molecules cm −3 (Kukui et al., 2012).In contrast, Palmer Station is exposed to oceanic air, producing ozone mixing ratios within the range of 9-20 ppbv with a mean OH concentration of about 1.0 × 10 5 molecules cm −3 (Jefferson et al., 1998).These observations suggest that NO x emissions increase oxidants at DDU about 2-fold for O 3 and more than 10-fold for OH compared to Palmer Station.Chemical transport models over the Antarctic continent show that NO x emitted from snow during summer increase O 3 and OH by a factor of 2 and 7, respectively, compared to estimations not including snow NO x emissions (Zatko et al., 2016).Additionally, 15 N depletion in nitrate starts from the beginning of September (See Supplement), which is consistent with previous measurement at DDU (Savarino et al., 2007), and supports the possibility that snow NO x emission happens in early spring at DDU.Thus, OH production is enhanced more efficiently than O 3 production in spring by snow NO x emission, possibly resulting in lower 17 O(nss-SO 2− 4 ) spring values.Another possible explanation is that hypohalous acids (HOX = HOCl and HOBr) act as important oxidants of SO 2 via the aqueous phase Reactions (R16) and (R17) in the marine boundary layer (Fogelman et al., 1989;von Glasow et al., 2002): This reaction is expected to produce sulfate with 17 O = 0 ‰ as all oxygen atoms of sulfate originate from water (Fogelman et al., 1989;Troy andMargerum, 1991: Yiin andMargerum, 1988), leading to lower 17 O(nss-SO 2− 4 ) values in the atmosphere.Indeed, unexpectedly low 17 O(nss-SO 2− 4 ) values have been observed in marine aerosols, which is possibly explained by a contribution of HOX oxidation of 33-50 % to total sulfate production in the marine boundary layer (Chen et al., 2016).Chen et al. (2016) estimated that a minimum concentration of gaseous HOX of 0.1 pptv could account for half of the sulfate production in the marine boundary layer.At DDU, yearround observations of gaseous inorganic bromine species ( ) revealed that maximum concentrations are observed in September, with values of 13.0 ± 6.5 ng m −3 (∼ 3.6 pptv) (Legrand et al., 2016b).Even if only one third of the Br * y corresponds to HOBr, as estimated using the model calculations by Legrand et al. (2016b) under conditions, then it is that HOX at DDU in spring is > 1 pptv; thus, HOX could play a significant role in production in spring.Likewise, Br * y concentration at DDU is at minimum in May, with values of 3.4 ± 1.0 ng m −3 (Legrand et al., 2016b) corresponding to less than one third of spring values.This would lead to a lower contribution of HOX to sulfate production in fall compared with spring.Hence, sulfate production via aqueous oxidation by HOX may explain the lower 17 O(SO 2− 4 ) values in spring relative to fall.
A change in the pH of the aqueous phase on the aerosol surfaces may also explain the spring/fall difference in 17 O(nss-SO 2− 4 ).Given that SO 2 oxidation by ozone in the aqueous phase is favored at high pH (> 5.5) (Seinfeld and Pandis, 2006), if the pH of aerosol droplets is higher in fall than in spring, then the relative importance of the SO 2 + O 3 reaction resulting in higher 17 O values would increase in fall.However, ion concentration analyses of aerosols collected at DDU exhibited higher alkalinity in spring than in fall (Jourdain and Legrand, 2002;Legrand et al., 2016b), which is inconsistent with this explanation.Hence, this process can be excluded from further consideration.
It should be noted that a smaller difference was observed between spring and fall 17 O(NO − 3 ) values, which would also be affected by the above two processes.Snow NO x emission would decrease 17 O(NO − 3 ) through depression of the contribution of O 3 oxidation relative to the other oxidation pathways by HO x and RO x , while halogen chemistry would lead to high 17 O(NO − 3 ) values through an oxygen atom transfer from O 3 to BrO and consequently to nitrate (Morin et al., 2007;Savarino et al., 2013).Although it is difficult to identify the precise processes involved, observations of 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values at an inland site (e.g., Concordia Station) would enable us to determine which process causes the spring/fall difference in the oxidation chemistry in the DDU atmosphere.If snow NO x emission is the source of low 17 O(nss-SO 2− 4 ) at DDU in spring, then 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) at an inland site would also exhibit lower values in spring than in fall.

Air mass origin analysis
Using observations of several oxidants at DDU and Concordia Station (e.g., Legrand et al., 2009Legrand et al., , 2016a;;Kukui et al., 2012Kukui et al., , 2014;;Grilli et al., 2013), it has been suggested that the oxidative capacity of the atmosphere at DDU is influenced by air masses transported from the East Antarctic plateau during katabatic wind outflows.The NO x emission from snowpack in inland Antarctica stimulates the ozone, HO x , RO x , and H 2 O 2 production through the enhanced NO x cycle.Thus, the atmosphere at DDU is enriched in all of those photochemical oxidants when air masses are from inland regions, compared to when air masses are from the ocean (Legrand et al., 2009).Therefore, we expected that 17 O values for sulfate and nitrate also depend on air mass origin.In Fig. 5, the 17 O values for sulfate and nitrate are given as a function of the time that air masses were over the continent during the five days' travel prior to arriving at DDU. Summer data show that the contact times of air masses with the continent varied 20 and 120 h (i.e., continental air and oceanic air were well mixed).Similarly, their 17 O values show insignificant variation, having low values of around 1 and 25 ‰ for sulfate and nitrate, respectively.This trend reflects two different phenomena: decreased 17 O values in summer because of the high contribution of photooxidants to atmospheric chemistry and increased import of oceanic air.In contrast, plots for other seasons show that the 17 O values exhibit high variation, although most of the air masses originate from the continent.It is important to note that this non-correlation does not mean that there is no link between 17 O values and air mass origin.The influence of air mass transport on oxidative capacity has been demonstrated in daily observations of oxidants (e.g., Legrand et al., 2009).Given that air masses at this site are from a variety of directions and are mixed together, this makes the interpretation of weekly averaged analyses more complicated.Hence, a non-significant correlation between 17 O values and air mass origin in weekly data could reflect a real lack of correlation or a too broad time resolution for these data.3 ) values are likely to reflect sunlight-driven changes in the relative importance of oxidation pathways; oxidation by HO x , RO x , and H 2 O 2 is increased during summer when the solar radiation enhances the production of those oxidants, whereas the relative contribution of the oxidation reaction transferring oxygen atoms of O 3 to sulfate and nitrate is increased during winter period.Interestingly, by comparing 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values to ozone mixing ratios, we found that the 17 O(nss-SO 2− 4 ) values in the spring months were lower than in fall months despite similar ozone levels for spring and fall, whereas there was no clear difference between 17 O(NO − 3 ) values.Possible explanations for the spring/fall differences for sulfate include (i) the low relative contribution of O 3 oxidation in spring induced by reactive nitrogen emissions from snowpack at inland sites being transported to coastal sites and (ii) the effects of SO 2 oxidation by hypohalous acids (HOCl and HOBr) enhanced in spring by the interaction of sea salt particles with photo-oxidants.Further observations of 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) in aerosols collected at Antarctic inland sites will help us identify the processes causing such different sulfate and nitrate formation in spring and fall.Nevertheless, the dataset of this study can be dedicated to atmospheric chemical transport models to improve the constraints on unique local oxidation chemistry at DDU.
O(O 3 ) values.Indeed, various 17 O(O 3 ) values ranging 25-35 ‰ are used for model calculations to reproduce the observed variations in 17 O(SO 2− 4 ) and 17 O(NO −

Figure 1 .
Figure 1.Comparison of 17 O values of sulfate for standard materials measured at the University of Washington (U.W.) and the Tokyo Institute of Technology (Tokyo tech.).
also supports this assumption.Since the δ 17 O and δ 18 O values of each quartz material used in this study are not known, the corrected δ 17 O and δ 18 O values of SO 2− 4 shown in the Supplement are unreliable and, therefore, we don't discuss these values.Note that those δ 17 O and δ 18 O values of SO 2− 4 are relative values to our O 2 reference gas.The precision of 17 O is typically better than ±0.2 ‰ based on replicate analyses of the standards.

Figure 2 .
Figure 2. Seasonal variations of concentrations (solid line) and 17 O values (circles) of sulfate (red), nitrate (blue), and ozone (purple) at the Dumont d'Urville Station during 2011. 17O values of ozone are shown as bulk (circle) and terminal (square) values.17O values of ozone include samples collected from January to April 2012.

Figure 3 .
Figure 3. Air mass pathways arriving at Dumont d'Urville Station during sampling periods in July 2011.

Figure 4 .
Figure 4. 17 O values of (a) sulfate and (b) nitrate as a function of ozone mixing ratios [O 3 ].Winter values were not taken into account in the calculation of slopes and intercepts.

Figure 5 .
Figure 5. Relationship between 17 O values of (a) sulfate and (b) nitrate with the time taken for air masses to pass over the Antarctic continent (i.e., air mass contact time with surface snow).
understanding of the factors influencing 17 O values of atmospheric sulfate and nitrate, seasonal variations of 17 O values of atmospheric sulfate, nitrate, and ozone were analyzed using the aerosol samples collected at DDU throughout 2011.Both 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values exhibited clear seasonal variations, with summer minima and winter maxima.In contrast, 17 O values of ozone showed limited variability throughout the year, indicating that 17 O(O 3 ) values do not significantly influence summer/winter trends in 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values.We hence, for the first time, demonstrated that 17 O(nss-SO 2− 4 ) and 17 O(NO − 3 ) values are the direct result of the oxidation chemistry of their precursors.The summer/winter trends of 17 O(nss-SO 2− 4 ) and 17 O(NO − ) values, but are mainly explained by the changes in oxidation pathways.To test this hypothesis, it is necessary to investigate the spatial and temporal variability of 17 O(O 3 ) values, which is expected to have the stable value, simultaneously with variations in 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values.Antarctica is a suitable site to test this hypothesis because of the clear seasonality in solar radiation, which is one of the main factors influencing 17 O(SO 2− 4 ) and 17 O(NO − 3 ) values through the changes in photochemical oxidant variations.In fact, several studies have reported the clear seasonal variations in 17 O(NO − ) values at Antarctica.In this study, we present the first simultaneous observations of 17 O values of sulfate, nitrate, and ozone at the coastal Antarctic site, Dumont d'Urville Station throughout 2011, to answer the key question about the controlling factors of 17 O(SO 2− 4 ) and 17 O(NO − . The 17 O(O 3 ) bulk values showed an insignificant variation, with a summer maximum of 28 ‰, a winter minimum of 23 ‰, and an annual mean 17 O(O 3 ) bulk value of 26 ± 1 ‰.

Table 1 .
Barkan and Luz (2003)) of sulfate and nitrate produced via each reaction pathway with oxidants. 17of sulfate are calculated based onSavarino et al. (2000).17Ovalues of nitrate refer to the estimate of the box model byMorin et al. (2011).17Ovalues of nitrate produced by BrONO 2 hydrolysis is assumed to be equal to the value for the NO 3 + RH pathway.Holt et al. (1983); b Vicars and Savarino (2014); c Savarino and Thiemens (1999); dBarkan and Luz (2003); e assumed based onBhattacharya et al. (2008); f a