South-hemispheric marine aerosol Hg and S isotope compositions reveal different oxidation pathways

: Particle-bound mercury (PBM) records the oxidation of elemental mercury, of which the main oxidation pathways (Br∙/Cl∙/OH∙/O 3 ) remain unclear, especially in the Southern Hemisphere. Here, we present latitudinal covariations of Hg and S-isotopic anomalies in cross-hemispheric marine aerosols that evidence an equator-to-poleward transition of Hg oxidants from OH∙/O 3 in tropics to Br∙/Cl∙ in polar regions highlighting thus the presence of distinct oxidation processes producing PBM. The correlations between Hg, S and O-isotopic compositions measured in PBM, sulfates and nitrates respectively within the aerosols highlight the implication of common oxidants in their formations at different latitudes. Our results open a new window to better quantify the present-day atmospheric Hg, S and N budgets and to evaluate the influences of aerosols on climate and ecosystems once the isotopic fractionations associated with each process have been determined.


INTRODUCTION
Gaseous elemental mercury (GEM, Hg 0 ) is the main atmospheric Hg species, with a lifetime up to 1 year and can therefore be transported worldwide [1].Hg 0 can be oxidized into gaseous oxidized mercury (GOM, Hg II ), a 10-day lifetime species which adsorbs readily on particles forming particle-bound Hg (PBM) [1].Both gaseous and particulate Hg II are transferred to terrestrial and marine ecosystems by rainfall and dry deposition.Upon deposition, Hg II is partially transformed into methylmercury that can bioaccumulate in aquatic organisms with well described impact on both wildlife and human health [2].Both PBM and GOM concentrations controlled by complex atmospheric oxidation and reduction reactions, and the chemical forms (Hg 0 , Hg II ) determine its atmospheric lifetime, deposition pathways and ecosystem loadings.However, many uncertainties persist in atmospheric mercury modeling, as the main atmospheric redox pathways of Hg remain unclear [3].Identified oxidants include OH•, O 3 , halogen species (mainly Br•) [4,5], and potentially NO 2 • and HO 2 • in a two-step oxidation mechanism initiated by Br• [6,7].Atmospheric models based on either type of Hg 0 oxidants (i.e., OH•/O 3 or Br•) show good agreements with observed Hg 0 concentration and Hg II wet deposition flux [6,8,9], but only after adjustment of terrestrial and marine Hg re-emissions.Recent computational and modeling studies suggest abundant photolysis of gaseous Hg-Br compounds back to Hg 0 , thereby competing with Hg II deposition [10,11].Hg II photolysis prolongs the atmospheric Hg lifetime up to 2 years, which is incompatible with observed Hg 0 variability in the planetary boundary layer [6,9], and suggests there are missing atmospheric oxidants.Atmospheric Hg redox mechanisms in global Hg transport and chemistry models are therefore ill-constrained, limiting our capacity for science-based Hg emission policy making.
Mercury stable isotopes could provide important information on the Hg 0 oxidation pathways because each of them should be associated with specific Hg-isotopic fractionations as demonstrated for halogen oxidants (e.g., Br•, Cl•) [12] and to be studied for some others including OH• and O 3 .In particular, Hg isotopes show mass-independent fractionation (MIF) mainly caused by a variety of radical reactions during photochemical processes.MIF of odd Hg isotopes (odd Hg-MIF, Δ 199 Hg and Δ 201 Hg, see Methods) can be induced by magnetic isotope effects (MIE) or nuclear volume effects (NVE), with most odd-MIF observed in nature produced during aqueous and gas phase photochemical reactions and abiotic reduction of Hg II [13][14][15].MIF of even Hg isotopes (even Hg-MIF), leading to distinct Δ 200 Hg and Δ 204 Hg signatures, is not fully understood yet and has been suggested to occur during Hg photo-oxidation at the tropopause [16].However, more recently, Sun et al. [12] suggested that the Hg 0 oxidation by halogen atoms (i.e., Br• and Cl•) would also produce both odd and even Hg-MIF, emphasizing the possible role of distinct Hg 0 oxidation pathway in the generation of MIF.Besides, significant insights into oxidative nitrate and sulfate formation, including quantifying the relative contributions of oxidants as OH•/O 3 or Br•/Cl•, i.e., the same as for Hg, have been made using sulfur (S) and/or oxygen (O) multiple isotopic compositions that also show MIF [17][18][19][20] (see Methods).For O-MIF, high values of Δ 17 O in nitrate or sulfate up to 36‰ and 8‰ [21] result from the oxidation of NO 2 or SO 2 by H 2 O 2 and O 3 which are characterized by a Δ 17 O varying from 1.3‰ to 2.2‰ [20] and 20‰ to 35‰ [22][23][24], respectively, while oxidants like OH• in the troposphere, BrO and O 2 +TMI produce sulfate with Δ 17 O ~0‰ [19,25,26].For S-MIF, the mechanisms suggested to be responsible for producing positive Δ 33 S in modern aerosols are stratospheric sulfate inputs into the troposphere [27,28] and/ or SO 2 oxidation by different oxidants (i.e., H 2 O 2 , O 3 , OH•, O 2 +TMI, BrO, etc.), while negative Δ 33 S results from a mechanism related to combustion or heterogeneous reaction [17,18,29].Despite the observation that PBM shows a significant relation with sulfate in aerosols [30] and shares similar oxidants (Supplementary Figure S1), S-Hg-O isotope systematics have never been coupled together to jointly understand the atmospheric oxidation budget and Hg redox cycling.
In this study, we target the Southern Hemisphere (SH) marine boundary layer, which provides an ideal place for studying atmospheric oxidation dynamics due to the absence of anthropogenic Hg and S point sources.We present new data for both Hg-(δ 202 Hg, Δ 199 Hg, Δ 201 Hg, Δ 200 Hg) and S-(δ 34 S, Δ 33 S and Δ 36 S) multiple isotopic compositions (see Methods) in marine aerosols collected shipboard during a cross-hemisphere round-trip cruise between Shanghai (China) and Antarctica (Figure 1, Supplementary Table S1).Coupled with O-isotopic compositions data from nitrate measured in the same set of samples [31], we show that the latitudinal isotopic gradients for Hg-and S-MIF in the SH indicate a shift in atmospheric oxidation pathway, and suggest that continental mineral dust plays an additional role in catalyzing Hg 0 and S oxidation.
During the positive excursions period, Δ 33 S is significantly correlated to Δ 200 Hg (P-value=0.04,Supplementary Figure S4c) but not to Δ 199 Hg (P-value=0.43,Supplementary Figure S4a), possibly suggesting that a similar reaction/mechanism produces both Δ 33 S and Δ 200 Hg.We also observe that SH aerosol data deviate from the Δ 200 Hg vs. Δ 199 Hg slope of 3.4 defined by atmospheric Hg data (Supplementary Figure S5).At present, no Hg mechanism at the Earth's surface is presently identified to generate Δ 200 Hg shifts.Our data suggest, independently from Δ 33 S, that we have detected production of Δ 200 Hg in the marine boundary layer.During the remaining cruise period (26/11/15 to the end of the cruise 10/04/16), Δ 33 S is significantly correlated to Δ 199 Hg (P-value=0.02,Supplementary Figure S4b) but not to Δ 200 Hg (P-value=0.93,Supplementary Figure S4d).Thus, at least two reactions (and MIF-generating mechanisms) are involved: one responsible for both Hg-MIF (i.e., Δ 199 Hg, Δ 201 Hg, Δ 200 Hg) and S-MIF (Δ 33 S<0.20‰) latitudinal gradients in the SH and another responsible for the concomitant Δ 33 S and Δ 200 Hg positive excursions off-shore Western Australia.

Latitudinal S-and Hg-MIF gradients caused by shift in oxidants
The symmetric pattern observed during the outward and return (O/R) trips, 4 months apart, in Figures 2 and 3 suggests that positive latitudinal gradients of Hg-MIF and Δ 33 S in the SH are a perennial feature.The gradients could result from a combination of control factors: a latitudinal isotope variability of Hg-and Semissions, a relation with the solar incidence or/and a latitudinal variation in the oxidation pathways.Varying contributions of two or more Hg-sources with latitude are not consistent with known natural sources (see Supplementary Figures S6 and S7).In particular, anthropogenic Hg and S contributions are expected to remain relatively low, consistent with the generally low NO 3 and SO 4 2 concentrations (except for the first three samples in the st-NH, see Supplementary Figure S8).Solar incidence also shows a latitudinal gradient suggesting a possible link to varying photochemical reactions.However, the solar insolation trends observed between November-December (outward trajectory) and March-April (return trajectory) in the Southern Hemisphere (0 to 90°S) are different with values varying from 450 to 500 W m −2 and 150 to 0 W m −2 , respectively [32], which would result in different intensity of photoreaction and thus different isotopic composition values.This is at odds with the similar Δ 201 Hg values (and gradients) determined between the outward and return trips (i.e., y=0.0083x and y=0.0113x, at a threshold of α=0.05%, t=0.241<t(16, α)=2.120 using a Welch's t-test respectively) or with the Δ 33 S values (and gradients) measured between the outward and return trips (i.e., y=0.0014x and y=0.0011x, at a threshold of α=0.05%, t=1.709<t (16, α) =2.120 using a Welch's t-test, respectively), ruling out the hypothesis.We suggest that in the SH marine boundary layer, the isotope signals of aerosol Hg and S may primarily be controlled by isotopic fractionation occurring during oxidation by oxidants such as O 3 , OH• and halogen compounds (Br• and Cl•) in the marine boundary layer.These oxidants have been shown to be important for the oxidation of SO 2 , species with a lifetime of ~2 days that result from the oxidation of DMS, the main S species emitted in the marine environment (up to 400 ppt vs. 200 ppt for SO 2 ) with a lifetime of 0.5 day [33], but also for Hg 0 [8,9], despite some theoretical doubts about the viability of O 3 and OH• [34].
A latitudinal change in oxidants is highlighted by the Δ 200 Hg varying from values within the range reported for the experimental oxidation by Br• and Cl• [12] (i.e., 0.10‰ to 0.10‰) to values which cannot be explained yet (Δ 200 Hg>0.10‰,Supplementary Figure S9), which we suggest might relate to an increase in the contribution of other oxidants in the Hg oxidation budget.Among the "other oxidants", OH• and O 3 are the most probable candidates as discussed above.Thus we will focus our discussion below on the possible role of OH• and O 3 in the Hg oxidation budget.However, we do not exclude the possibility that other unknown oxidants may also play a role.We use the GEOS-Chem atmospheric Hg chemistry and transport ).Note that the scale is different for each oxidant.
model to estimate the annual, spatial contributions of four oxidants, OH•, O 3 , Br•, and Cl•, to Hg 0 oxidation in the marine boundary layer and lower troposphere (0-3 km; see Methods).Our results show latitudinal variation of Hg 0 oxidation dominated by OH• and O 3 accounting for ~80% from 0 to 40°S, to an oxidation dominated by Br• accounting for ~75% of oxidation from 40°S to 80°S (Figure 4).Although the kinetic rate for the Hg 0 oxidation by O 3 estimated to be 10 −19 and 10 −20 cm 3 /molecules [35,36] is the slowest among these oxidants [37,38], oxidation by O 3 needs to be considered given its relatively high concentrations compared with OH• or Br• (ppbv [39] vs. pptv [40] level, respectively), leading to significant amounts of Hg II oxidized by O 3 , i.e., maximum Hg II produced by OH• and O 3 oxidation pathways of 5 and 1 mg yr −1 , respectively (Figure 4).Thus, the presence of different oxidation pathways could account for the Hg isotopic compositions and budget.This argument is in line with the recent study from Saiz-Lopez et al. [41], who proposed that at least one oxidation pathways responsible for the formation of PBM in the atmosphere is missing based on modeling study.It is however worth noting that aqueous photoreduction may also change the odd-MIF and thus could in theory explain the Δ 199 Hg vs. Δ 201 Hg slope of 0.93±0.1.However, photoreduction alone is unable to account for the variation of Δ 200 Hg up to 0.20‰ measured during the remaining cruise period (that is not related to the Δ 33 S from 26/11/15 to the end of the cruise 10/04/16), because photoreduction has not been shown to trigger any even-MIF as reported in previous studies [13,15].Aqueous abiotic reduction of Hg II can produce very small negative even-MIF in the remaining Hg(II) via NVE [42].However, Δ 200 Hg caused by NVE is proportional to δ 202 Hg and is typically <0.10‰ unless δ 202 Hg is unrealistically large (e.g., >10‰) [42].Thus NVE cannot explain the up to 0.20‰ variation in Δ 200 Hg.For the S system, SO 4 2 production by progressive oxidation of marine DMS has typical values of δ 34 S ~16‰ and Δ 33 S ~0.05‰ [43].Similar spatial distribution and variation of the annual mean concentration of gas-phase BrO, OH• and O 3 involved in DMS oxidation in the SH have been reported with oxidation dominated by OH•/O 3 up to 60% at latitude varying from 0°to 40°S to an oxidation dominated by BrO up to 70% at latitude varying from 40°S to 80°S [43].Although S isotope fractionation associated with DMS oxidation to SO 2 has not been studied yet, the transition from dominant DMS oxidation by O 3 /OH• to halogen compound is globally consistent with the GEOS-Chem results for Hg (Figure 4) [43].Although measurement of O-MIF in sulfate from the same set of aerosols samples, which have been shown to be significantly correlated to the S-MIF in aerosols [28], would have brought additional information on the possible implication of different oxidation pathways, this data could not be obtained due to limited sample.However, the analyses of Δ 17 O in NO 3 for the same set of samples in the SH [31] also show a latitudinal gradient with lower Δ 17 O in the low latitudes and higher Δ 17 O in polar region (Supplementary Figure S10, [31]), signing the transition from NO 2 oxidation by OH•/O 3 at the equator to oxidation by BrO near Antarctica [31].Significant correlations between i) the Δ 199 Hg with the Δ 17 O (ρ(26)=−0.4132,P-value=0.03,Supplementary Figure S11a) and ii) Δ 33 S with the Δ 17 O (ρ(26)=−0.5050,P-value=0.01,Supplementary Figure S11b) suggest that these proxies (sulfate, nitrate and PBM) might be controlled by the same oxidants and that latitudinal oxidation gradient (pathways) for different species might be controlled by the concentration of different oxidants despite having distinct chemical reactions.Our data converge so that Hg 0 oxidation by OH•/O 3 might be a significant perennial process, dominant in the sub-tropics, and responsible for producing odd Hg-MIF, and possibly even Hg-MIF, with Δ 200 Hg<0.20‰observed in SH aerosols.Future experimental Hg isotope studies are required to demonstrate even Hg-MIF together with experimental DMS oxidation by Natl Sci Open, 2022, Vol.1, 20220014 different oxidant to determine their isotopic fractionation.

Heterogeneous photoreactions on mineral dust might trigger even Hg-MIF
Super-imposed on the latitudinal gradients of both Hg-and S-MIF, which we attribute to the transition from OH•/O 3 oxidation to halogen oxidation, are the positive Δ 200 Hg and Δ 33 S excursions of 0.30‰ and 0.40‰, respectively, that occur off-shore West-Australia (Figure 3).The restriction of these positive Δ 33 S and Δ 200 Hg excursions only during the outward trip highlights a sporadic mechanism.The mechanism triggering positive Δ 200 Hg remains unclear [44], and significant insights could be gained from S-isotopes of the same samples.Both Δ 200 Hg and Δ 33 S are generally thought to be associated with stratospheric inputs [16,45,46] that have been reported intensively in this area, where stratospheric sulfate possibly reaches 1 μg m −3 [47].Considering that the concentration of sulfate could reach 4 μg m −3 in our study, this could represent an injection of 25% of stratospheric sulfate in our samples, which could account for the highest Δ 33 S of 0.44‰ considering that stratospheric sulfate would be characterized by a maximum Δ 33 S of 10‰ [48].However, this is inconsistent with our data which shows that samples characterized by both the highest Δ 33 S up to 0.44‰ and Δ 200 Hg up to 0.17‰ are composed of 90% of sea-salt sulfate (Supplementary Information).Moreover, if stratospheric injection were the sole mechanism responsible for the high Δ 33 S and Δ 200 Hg, high values of odd-MIF would also be expected due to the high contribution of photoreduction, which is not observed as these samples are characterized by low values of −0.07‰ and −0.10‰ for the odd-MIF.Furthermore, three-day HYSPLIT backward trajectory analysis for each sample, which is usually performed to discuss the S-MIF origin [45,49], shows that none originate from altitudes higher than 2000 m for initial trajectory release heights of 10 or 500 m, respectively (Supplementary Figure S12).Together with the fact that stratospheric-tropospheric mass flux exchanges are low in the tropics (30°N-30°S) with values varying from 0 to 20 kg s −1 km −2 [50] based on a lapse-rate tropopause method (comparatively, the highest mass flux is 600 kg s −1 km −2 in 30°N-60°N [50]), this again suggests that upper tropospheric or stratospheric Hg and S inputs would not be the sole mechanism responsible for the Δ 200 Hg and Δ 33 S excursions.To explain the S-MIF in sulfate aerosols varying from -0.60‰ to -0.50‰, photooxidation of SO 2 in the presence of mineral dust was recently suggested [29,51].Although the implicated mechanism remains unconstrained yet, this hypothesis may account for both the higher SO / Ca 4 2 2+ enrichment factor used as a geochemical tracer for dust (see Supplementary Information) and the difference of Δ 33 S between the O/R trip between Shanghai and Australia, as dust storms frequently occur from September to December in Australia [52].The dust input hypothesis is consistent with i) the forward trajectories of dust particles concentration using HYSPLIT during this cruise which shows that dust particle paths may meet the sampling path only during the outward trip (Supplementary Figure S13); and ii) both positive Δ 33 S and Δ 36 S up to 0.44‰ and 1.21‰, respectively, measured in these samples that could only, to date, be explained by photooxidation [53] (Supplementary Figure S3).By analogy, we suggest that the occurrence of positive Δ 200 Hg excursions up to 0.30‰ may also result from Hg 0 photo-oxidation catalyzed by mineral dust surfaces.This hypothesis is consistent with Chen et al. [16] who suggest that oxidation on aerosol surfaces could potentially promote the transformation of Hg 0 and produce positive Δ 200 Hg.Although speculative, our findings suggest the possibility of even Hg-MIF produced at the Earth's surface, and should stimulate new experiments involving heterogeneous Hg transformations on aerosol and mineral surfaces to constrain the potential mechanisms.

Atmospheric Hg transformations over Antarctica
We observe negative Δ 199 Hg and Δ 201 Hg ranging from −0.50‰ to −0.30‰ during the ship's stay in Antarctica (Figure 3).These are associated with a strong increase in Hg concentration up to 140 pg m −3 (Supplementary Figure S8), while the concentration during the rest of the cruise is ~4 pg m −3 , consistent with literature data (2 to 8 pg m −3 ) [54,55].These variations are associated with variations in Δ 200 Hg and Δ 33 S close to 0‰ (from −0.01‰ to 0.08‰ and 0.00‰ to 0.06‰, respectively) highlighting again, other types of atmospheric Hg redox mechanisms.Aerosol Hg II concentration peaks and negative odd Hg-MIF shift during the summer stay in Antarctica are unlikely to reflect higher Hg re-emission from seawater, since marine Hg is characterized by zero to positive Δ 199 Hg up to 0.40‰ [43].This is supported by the low Na + and Cl − concentrations in these samples (Supplementary Figure S8) and HYSPLIT backward trajectories showing that the associated air masses are mainly stagnant over the Antarctic plateau (Supplementary Figure S12).Instead, the Hg II originally deposited on snow could be photoreduced into Hg 0 , being later re-oxidized in the atmosphere and finally enriched on particles on the Antarctic plateau.Polar environments are unique with respect to the abundant photochemical release of Br from sea-ice and snow, which leads to intense Hg 0 oxidation especially during the springtime atmospheric mercury depletion event (AMDE) [5].AMDEs occur both in the Arctic and Antarctica and deposit large amounts of Hg II to surface snow, consistent with the intense Hg 0 oxidation reported on the Antarctic plateau [56,57] following the accumulation of oxidants within the shallow boundary layer during summer [57].A significant fraction of deposited Hg II in snowpack is then subsequently photoreduced and re-emitted back to the atmosphere [58].This hypothesis is consistent with the fact that Hg II -photoreduction in snow is associated with negative Δ 199 Hg down to −5.50‰ in the residual Hg II pool, which possibly could explain the initial negative Δ 199 Hg (i.e., −0.50‰) measured in Antarctica aerosols.Moreover, successive oxidation and photoreduction could account for the Δ 199 Hg shift observed from −0.50‰ to −0.30‰ in samples collected in Antarctica consistent with available evidence that photoreduction produces higher Δ 199 Hg in Hg 0 and no change in Δ 200 Hg [59] (Figure 3).Overall, these results are consistent with the known model of cyclical Hg oxidation and reduction in Antarctica during spring and summertime [60].

Implications for the oxidizing potential of the atmosphere
The latitudinal gradient of Hg isotope MIF in the aerosols collected along the cross-hemisphere cruise (30°N to 60°S) demonstrates that the two-step oxidation initiated by Br• cannot be the sole process responsible for Hg 0 oxidation in the SH marine boundary layer, emphasizing the need to consider additional oxidation pathways.This result is also consistent with the recent study from Saiz-Lopez et al. [41].Modeled oxidation of S and Hg (also N) suggest an important shift in primary oxidants from the SH sub-tropics (OH•, O 3 ) to midlatitudes and polar regions (Br•).Mass-independent signatures of S, O, and Hg isotopes in aerosols covary with the shift in oxidants, showing higher MIF in the subtropics which we relate to an increase in Hg oxidation by OH• and O 3 .Previous global Hg modeling studies have opposed OH•, O 3 and Br as primary Hg oxidants [6,8,9].Our study suggests independently from the model studies that both pathways operate in the different regions of the marine boundary layer and should be taken into account simultaneously, although future experimental isotopic studies need to be performed to support it and to quantitatively determine the Natl Sci Open, 2022, Vol.1, 20220014 contribution of each oxidation pathway.We report additional S-MIF and even Hg-MIF signatures that might reflect the contribution of a reaction catalyzed by mineral dust emitted from the Australian continent.Heterogeneous oxidation chemistry on aerosol surfaces may provide the missing oxidation capacity observed in recent global atmospheric Hg models.Evidence for Hg 0 oxidation by OH•/O 3 could be important in regions that are characterized by high concentrations of O 3 and might explain the positive relation between the Particles-Bound Mercury concentration and PM 2.5 concentration (i.e., mainly formed by physical and chemical processes in the atmosphere) observed in urban environment during pollution events [61].These oxidation pathways reduce Hg 0 lifetime and increase Hg II deposition rate locally, changing both Hg II and Hg 0 fluxes.This ultimately leads to higher risk for human health as it would increase Hg II deposition rate, for example, in tropical Asia (Figure 4) where fisheries catch dominates [62].Further studies on combined MIF of O, S and Hg isotopes may also help understand a variety of scientific issues that could not be solved by single isotope system, such as the O-MIF of nitrate in Antarctica which remains debated [63,64], the origin of both sulfate and nitrate formed during polluted events in urban environment, and the evolution of Earth's atmosphere over geological times.

Definition of MIF
The notation Δ is applied for both S and Hg isotopes as follows: Δ 33

Sample collection and processing
Aerosols were collected along the cruise track from Shanghai, China (31.32°N, 121.69°E) to East Antarctica (Zhongshan Station; 69.37°S, 76.37°E) during the 32nd Chinese Antarctic research expedition from November 2015 to April 2016.The aerosols were collected using a high volume air sampler (HVAS), which was placed on the top deck of the icebreaker RV Xuelong, about 25 m above the sea surface.The HVAS was controlled by wind direction and wind speed sensors, and it was operated only when the incoming wind was from a sector about 90°left and right on the ship's forward path, with the wind speed greater than 1.5 m s −1 , to avoid contamination of the ship emissions.The samples were collected on Whatman G653 glass-fiber filters (8×10in), which were prebaked at 550°C for 24 h and which have been used for both Hg, S and O isotopic compositions [29,44,49,61,[64][65][66][67][68].Although this is the conventional sampling method for measuring the Hg isotopic compositions, several studies have shown that those methods lead to biases in PBM concentrations [69,70], underestimating their concentration.However, we argue that the MIF of PBM would not be significantly modified by these bias, because Hg MIF is primarily caused by photochemical processes.In general, the sampling durations were generally 24-72 h for each sample, with a flow rate of about 1.0 m 3 min −1 , thus leading to the sampling air volume of 1500-4000 m 3 .In total, 42 aerosol samples were collected along the cruise path (Figure 1).All samples were kept in clean opaque polythene bags, preserved in insulated cabinets, and transported to China at -20°C.
In the laboratory, three quarters of each glass fiber filters were cut into pieces using pre-cleaned scissors that were rinsed between samples, placed in 100 mL Milli-Q water (18.2MΩ), ultrasonicated for 40 min and leached for 24 h under shaking at 120 r min −1 .The extract solutions were then filtered through 0.22 μm PTFE filters for chemical ion concentration and isotopic composition (sulphate and nitrate) measurements.The remaining one quarter of filter was used for Hg isotope analysis.

Chemical analysis
Chemical ion concentration in sample solutions was determined using an ion chromatograph (ICS3000, Dionex, USA) following previous work [71].In brief, the column used for cation analysis (Na + , NH 4 + , K + , Mg 2+ and Ca 2+ ) was a Dionex column CS12 (2 mm×250 mm), with a guard column CG12 (2 mm×50 mm); while the anions (MSA, Cl − , NO 3 and SO 4 2 ) were analyzed using a Dionex column AS11 (2 mm×250 mm) with a guard column AG11 (2 mm×50 mm).The eluents for cations and anions were MSA and potassium hydroxide (KOH), respectively.The relative standard deviation of ion concentrations in replicate samples (n=5) was generally <5%.Concentrations of Hg were measured by a cold vapor atomic fluorescence spectroscopy (CVAFS, Tekran 2500) after reducing an aliquot of the pre-concentrated Hg II into Hg 0 using SnCl 2 .The Hg 0 is then collected on a gold-coated bead trap and analyzed [72].

S multi-isotope analysis
Part of the extracted solution was inserted into a reaction vessel heated at 180°C with 20 mL of Thode solution, a mixture of hydrochloric, hydroiodic and hypophosphorous acids [73], for 1.3 h to quantitatively reduce sulfate into H 2 S. The formed gases were purged from the vessel using nitrogen gas, bubbled through deionized water and subsequently passed through a 0.3 mol L −1 silver nitrate (AgNO 3 ) solution to form silver sulfide (Ag 2 S).This solid Ag 2 S was then rinsed twice with Millipore water and dried at 70°C overnight.Ag 2 S was then loaded into an aluminum foil, weighted and degassed under vacuum.
Ag 2 S was subsequently converted to SF 6 by reacting with approximately 200 Torr of excess fluorine in a nickel bomb at 250°C (1 Torr=1.333×10 2 Pa).The produced SF 6 was purified using both cryogenic techniques and gas chromatography, quantified and subsequently analyzed by dual inlet isotope ratio mass spectrometry (Thermo-Fisher MAT-253) where m/z=127, 128, 129 and 131 ion beams were monitored.
The δ 34 S-values were measured against our in-house SF 6 tank that had been previously calibrated with respect to the IAEA-S1 international standards and expressed versus V-CDT assuming a δ 34 S S1 =−0.3‰ vs. V-Natl Sci Open, 2022, Vol.1,20220014 CDT isotope composition.To express our Δ 33 S and Δ 36 S data with respect to V-CDT, we anchored our data using CDT-data measured previously in the laboratory following Defouilloy et al. [74].No further corrections were carried out, other than normalization of the data to CDT.Δ 33 S and Δ 36 S IAEA-standards were within values reported elsewhere [74][75][76].Our analysis (n=5) of IAEA-S1 standard yielded: δ 34 S=−0.33‰±0.02‰, Δ 33 S=0.074‰±0.010‰ and Δ 36 S=−0.809‰±0.128‰ vs. CDT.Analyses of the international sulfate standard NBS-127 was also performed and gave a δ 34 S of 20.8‰±0.4‰(2σ; n=12), consistent with the 20.3‰±0.4‰value reported by the IAEA.Field blank samples (n=2) were analyzed and generated no sulfate when we followed the chemistry method described below.

Hg multi-isotope analysis
The remaining 1/4 filters have been combusted using a dual stage combustion protocol to pre-concentrate the Hg.This protocol is detailed in Huang et al. [72].Briefly, the filter is introduced in a quartz tube, and inserted in the first tube furnace, which rapidly ramps the temperature to 950°C for 3.5 h, while the second furnace is maintained at a constant temperature of 1000°C to pyrolize all combustion products to CO 2 , H 2 O, SO 2 , and Hg 0 [77].The combustion products are purged by Hg-free O 2 and bubbled in a 5 mL acid trap consisting of a mixture of HNO 3 , HCl and water (2:4:9) [72,77].The acid trapping solution is then transferred to a precleaned glass bottle, diluted with milli-Q water and 50 μL of 0.2 mol L −1 BrCl.

LOWESS and statistics
The LOWESS smoothing has been calculated by MATLAB using a linear polynomial function with a span of 25%.
In the following, the significance of each correlation is calculated using a Spearman Correlation as this non-parametrical test does not carry any assumption about the data distribution and is not sensitive to outliers.

Hysplit analysis
To investigate relationships between the origins of the air masses and the isotopic compositions of S and Hg measured in marine aerosols, 72 h backward trajectories at 10 and 500 m height have been modeled using HYSPLIT-Hybrid Single Particles Lagrangian Integrated Trajectory-for each sample.The model used NCEP-NCAR reanalysis data fields with a resolution of 0.5×0.5.Backward trajectories for the samples are then incorporated into a map generated by generic mapping tools.Global Relief data come from ETOPO2 (last access: 06/05/19) [85].In addition, 72 h foreward trajectories for dust emission sources have been modeled using HYSPLIT.Coordinates of Australian deserts were compiled from the NOAA and Natural Earth (last access: 06/05/19 https://www.arl.noaa.gov/hysplit/dust/and https://www.naturalearthdata.com/downloads/).The simulation considered an arbitrary particle released by cycle of 100 and an emission rate of 61Tg/year [86]).

GEOS-Chem simulation
We use a global three-dimensional atmospheric chemistry and transport model GEOS-Chem (version 12.1.0)to simulate the Hg redox chemistry (www.geos-chem.org)following Horowitz et al. [6].It has a horizontal resolution of 4°latitude ×5°longitude with 47 vertical layers.The model is driven by GEOS-FP assimilated meteorological data from the NASA Global Modeling and Assimilation Office.The anthropogenic emissions are from Zhang et al. [87] with land and ocean re-emissions from Selin et al. [88] and Zhang et al. [89].The model induces Hg redox chemistry summarized by Horowitz et al. [6], in which Hg 0 is oxidized to Hg II in a two-stage process including an initial oxidation by halogen atoms (Br and Cl) and secondary oxidation by Br, Cl, HO 2 , OH, NO 2 , ClO and BrO.In addition, we include the oxidation of Hg 0 by OH and O 3 following Selin et al. [88].The reduction of Hg II to Hg 0 is simulated as photolysis of aqueous-phase Hg II -organic complexes in aerosols and clouds.The gas-particle partitioning between Hg II and Hg p is simulated as a thermodynamic equilibrium function of local temperature and fine particulate matter concentrations (PM 2.5 ) following Amos et al. [90].The model is run for 2014-2016 with the result of first year discarded as spin-up time.

Figure 1
Figure 1 Map showing the sampling location during the round-trip cruise from November 2015 to April 2016.

Figure 2 Figure 3
Figure 2 Variation of (A) Δ 199 Hg, (B) Δ 200 Hg, (C) Δ 201 Hg, (D) Δ 33 S during both outward and return (O/R) cruises as a function of latitude from the st-NH to the SH.A linear regression is applied for samples in the st-NH and SH.The Spearman correlation is reported by the ρ(df) notation for both st-NH and SH which include both O/R trip.df is the degree of freedom and the significance is reported by the P-value.