NO 2-initiated multiphase oxidation of SO 2 by O 2 on CaCO 3 particles

The reaction of SO2 with NO2 on the surface of aerosol particles has been suggested to be important in sulfate formation during severe air pollution episodes in China. However, we found that the direct oxidation of SO2 by NO2 was slow and might not be the main reason for sulfate formation in ambient air. In this study, we investigated the multiphase reaction of SO2 with an O2 /NO2 mixture on single CaCO3 particles using Micro-Raman spectroscopy. The reaction converted the CaCO3 particle to a Ca(NO3)2 droplet, with CaSO4 q2H2O solid particles embedded in it, which constituted a significant fraction of the droplet volume at the end of the reaction. The reactive uptake coefficient of SO2 for sulfate formation was on the order of 10−5, which was higher than that for the multiphase reaction of SO2 directly with NO2 by 2–3 orders of magnitude. According to our observations and the literature, we found that in the multiphase reaction of SO2 with the O2 /NO2 mixture, O2 was the main oxidant of SO2 and was necessary for radical chain propagation. NO2 acted as the initiator of radical formation, but not as the main oxidant. The synergy of NO2 and O2 resulted in much faster sulfate formation than the sum of the reaction rates with NO2 and with O2 alone. We estimated that the multiphase oxidation of SO2 by O2 initiated by NO2 could be an important source of sulfate and a sink of SO2, based on the calculated lifetime of SO2 regarding the loss through the multiphase reaction versus the loss through the gas-phase reaction with OH radicals. Parameterization of the reactive uptake coefficient of the reaction observed in our laboratory for further model simulation is needed, as well as an integrated assessment based on field observations, laboratory study results, and model simulations to evaluate the importance of the reaction in ambient air during severe air pollution episodes, especially in China.


Introduction
It has been suggested that multiphase or heterogeneous oxidation of SO 2 potentially plays an important role in sulfate formation in the atmosphere (Seinfeld and Pandis, 2006).During the severe pollution episodes that occur frequently in China, high sulfate concentrations cannot be explained by the gas-phase oxidation of SO 2 and its well known aqueous chemistry (Zheng et al., 2015a;Cheng et al., 2016), highlighting the role of underappreciated heterogeneous oxidation or multiphase pathways.
Recently, the multiphase oxidation of SO 2 by NO 2 has been introduced in air quality model simulations to explain the discrepancy between the modeled and observed sulfate concentration during severe pollution episodes in China (Cheng et al., 2016;Gao et al., 2016;Wang et al., 2016;Xue et al., 2016), despite the uncertainties in the kinetic parameters for SO 2 oxidation and in the pH value of aerosol particles in China (Wang et al., 2016;Cheng et al., 2016;Liu et al., 2017;Guo et al., 2017).However, according to our recently published results (Zhao et al., 2018), the direct oxidation of SO 2 by NO 2 could not contribute significantly to sulfate formation in the atmosphere because the reactive uptake coefficient of SO 2 for sulfate formation due to direct oxidation by NO 2 is very low (∼ 10 −8 ).
Published by Copernicus Publications on behalf of the European Geosciences Union.
Although the contribution of the direct oxidation of SO 2 by NO 2 to sulfate formation is not significant, NO 2 may be involved in other oxidation pathways of SO 2 .It has been reported that the reaction of NO 2 with SO 2− 3 and HSO − 3 in the bulk aqueous phase can form the SO q − 3 radical, which can further react with O 2 and produce a series of radicals that oxidize S(IV) species (Littlejohn et al., 1993).The reaction pathway may result in a fast SO 2 oxidation due to the potential synergy of NO 2 and O 2 .
Despite such a reaction mechanism for SO 2 oxidation being proposed, its role in SO 2 oxidation in the ambient atmosphere is not well established.Most previous studies have focused on the direct reaction of SO 2 with NO 2 , including the determination of its rate constant (Lee and Schwartz, 1983;Clifton et al., 1988;Shen and Rochelle, 1998;Spindler et al., 2003;Nash, 1979;Huie and Neta, 1986).According to the reaction products and their reported yields (Lee and Schwartz, 1983;Clifton et al., 1988), the overall reaction equations of the direct reaction of SO 2 with NO 2 are as follows: 2NO 2 (aq) and the reactions are proposed to proceed via NO 2 -S(IV) adduct complexes (Clifton et al., 1988).
NO 2 (aq) + SO 2− 3 (aq) → [NO 2 −SO 3 ] 2− (aq) (R3) NO 2 (aq) + [NO 2 −SO 3 ] 2− (aq) Additionally, NO 2 -S(IV) adduct complexes may decompose as follows (Spindler et al., 2003): However, studies of the oxidation rate of SO 2 at the O 2 concentrations relevant to the ambient atmosphere and the potential influence of the synergy of NO 2 and O 2 on the oxidation rate are very limited (Turšič et al., 2001;He et al., 2014), except for a few studies which investigated SO 2 oxidation in the presence of NO 2 as well as O 2 (Littlejohn et al., 1993;Shen and Rochelle, 1998;Santachiara et al., 1990).Moreover, previous studies have mainly focused on the reaction in bulk solution and only few studies have investigated the oxidation of SO 2 by NO 2 on aerosol particles (Santachiara et al., 1990(Santachiara et al., , 1993)).On aerosol particles, water activity, pH, ionic strength, the presence of other compounds or ions, and the role of the particle surface are different from dilute bulk solution and may affect the reaction process and reaction rate.Therefore, further studies of the multiphase reaction of SO 2 with O 2 / NO 2 mixtures on aerosol particles are required to determine the kinetic parameters and the mechanism of the reaction.
In this study, we investigated the multiphase reaction of SO 2 with O 2 in the presence of NO 2 on CaCO 3 particles.We quantified the reactive uptake coefficient of SO 2 due to the reaction with an O 2 / NO 2 / H 2 O mixture.Based on our observations and the existing literature, we further discussed the reaction mechanism.Furthermore, we estimated the role of the multiphase oxidation of SO 2 by O 2 in the presence of NO 2 in the atmosphere.

Experimental
The experiments were conducted using a flow reaction system and the setup is shown in Fig. S1 in the Supplement.The experimental setup and procedure used have been described in detail in previous studies (Zhao et al., 2018(Zhao et al., , 2011;;Liu et al., 2008).A gas mixture of NO 2 , SO 2 , O 2 , N 2 , and water vapor reacted with particles deposited on a substrate in the flow reaction cell.The concentrations of SO 2 and NO 2 were controlled using mass flow controllers by varying the flow rates of SO 2 (2000 ppm in high purity N 2 , National Institute of Metrology P.R. China), NO 2 (1000 ppm in high purity N 2 , Messer, Germany), and synthetic air (20 % O 2 (high purity grade: 99.999 %, Beijing Haikeyuanchang Practical Gas Co., Ltd.) and 80 % N 2 (high purity grade: 99.999 %, Beijing Haikeyuanchang Practical Gas Co., Ltd.)).Relative humidity (RH) was controlled by regulating the flow rates of reactant gases, dry synthetic air, and humidified synthetic air.Humidified synthetic air was prepared by bubbling synthetic air through fritted glass in water.In some experiments, the O 2 concentrations were varied by regulating the mixing ratios of O 2 and N 2 to investigate the effect of O 2 .SO 2 / O 2 / NO 2 / H 2 O mixtures flowed through the reaction cell and reacted with individual stationary CaCO 3 particles, which were deposited on a Teflon ® FEP film substrate annealed to a silicon wafer.RH and temperature were measured using a hygrometer (HMT100, Vaisala, Vantaa, Finland) at the exit of the reaction cell.Additionally, temperature was measured using another small temperature sensor (Pt 100, 1/3 DIN B, Heraeus, Hanau, Germany) in the reaction cell.All the experiments were conducted at 298 ± 0.5 K.The experiments were conducted under two RHs (72 and 82 %) at 75 ppm SO 2 and 75 ppm NO 2 .
During the reaction, particles were monitored in situ via a glass window on the top of the reaction cell using a Micro-Raman spectrometer (LabRam HR800, HORIBA Jobin Yvon, Kyoto, Japan) to obtain microscopic images and Raman spectra.A 514 nm excitation laser was used, and back-scattering Raman signals were detected.The details of the instrument are described elsewhere (Liu et al., 2008;Zhao et al., 2011).Because the particles were larger than the laser spot in this study (∼ 1.5 µm), confocal Raman mapping was used to measure the spectra at different locations on a particle to obtain the chemical information of the entire particle.The mapping area was rectangular and was slightly larger than the particle, with mapping steps of 1 × 1 µm.Raman spectra in the range of 800-3900 cm −1 were acquired with an exposure time of 1 s for each mapping point.Raman spectra were analyzed using Labspec 5 software (HORIBA Jobin Yvon).Raman peaks were fitted to Gaussian-Lorentzian functions to obtain peak positions and peak areas at different locations on the particle.The peak areas were then added together to obtain the peak area for the entire particle.
Particles of CaCO 3 (98 %, Sigma-Aldrich, USA), with average diameters of about 7-10 µm, as specified by the supplier, were used in the experiments.The CaCO 3 particles were rhombohedron crystals; X-ray diffraction analysis indicated that they were calcite (Fig. S2).Individual particles were prepared by dripping a dilute CaCO 3 suspended solution onto Teflon ® FEP film using a pipette and then drying the sample in an oven at 80 • C for 10 h.
The amount of CaSO 4 as a reaction product was quantified based on Raman peak areas and particle sizes.The details of the method are described in our previous study (Zhao et al., 2018).Briefly, the amount of reaction product CaSO 4 formed was determined as a function of time using Raman peak areas.Raman peak areas were converted to the amount of compound formed using a calibration curve obtained from pure CaSO 4 particles of different sizes, which were determined according to microscopic images.The reaction rate, i.e., the sulfate production rate, was derived from the amount of sulfate formed as a function of time.The reactive uptake coefficient of SO 2 for sulfate formation (γ ) was further determined from the reaction rate and collision rate of SO 2 on the surface of a single particle.
where R is the gas constant, T is temperature, M SO 2 is the molecular weight of SO 2 , c is the mean molecular velocity of SO 2 , A s is the surface area of an individual particle, and Z is the collision rate of SO 2 on the surface of a particle.{SO 2− 4 } indicates the amount of sulfate in the particle phase in moles.The average reaction rate and surface area of particles during the multiphase reaction period were used to derive the reac-tive uptake coefficient.The period was chosen to start after the induction period when ∼ 10 % of the final sulfate was formed.[SO 2 ] indicates the concentration of SO 2 in the gas phase.
The influence of gas-phase diffusion on reactive uptake was evaluated using the resistor model described by Davidovits et al. (2006) and references therein, as well as using the gas-phase diffusion correction factor for a reactive uptake coefficient according to the method described by Pöschl et al. (2007).The reactive uptake of SO 2 was found to not be limited by gas-phase diffusion (see details in the Supplement, Sect.S1).
In addition, we conducted experiments of the reaction SO 2 with only O 2 on both CaCO 3 solid particles and internally mixed CaCO 3 / Ca(NO 3 ) 2 particles (with CaCO 3 embedded in Ca(NO 3 ) 2 droplets), while keeping other conditions the same as the reaction of SO 2 with an O 2 / NO 2 mixture.These experiments of the multiphase oxidation of SO 2 by O 2 can help determine the role of NO 2 in the reaction of SO 2 with an O 2 / NO 2 mixture.

Reaction products and changes in particle morphology
Figure 1 shows the Raman spectra of a CaCO 3 particle during the multiphase reaction of SO 2 with O 2 / NO 2 / H 2 O on its surface.The peak at 1087 cm −1 was assigned to the symmetric stretching of carbonate (ν s (CO 2− 3 )) (Nakamoto, 1997).During the reaction, the peak at 1087 cm −1 decreased continuously and finally disappeared as new peaks were observed.The peak at 1050 cm −1 was assigned to the symmetric stretching of nitrate (ν s (NO − 3 )).The peaks at 1010 and 1136 cm −1 were assigned to the symmetric stretching (ν s (SO 2− 4 )) and asymmetric stretching (ν as (SO 2− 4 )) of sulfate in gypsum (CaSO 4 q 2H 2 O), respectively (Sarma et al., 1998).In addition, after the reaction, a broad envelope in the range of 2800-3800 cm −1 assigned to the stretching of the OH bond in water molecules was observed.Above this envelope, there were two peaks at 3408 and 3497 cm −1 , which were assigned to OH bond stretching in crystallization water of CaSO 4 q 2H 2 O (Sarma et al., 1998;Ma et al., 2013).
During the multiphase reaction with the SO 2 / O 2 / NO 2 / H 2 O mixture, the CaCO 3 particles displayed a remarkable change in morphology.The original CaCO 3 particle was a rhombohedron crystal (Fig. 2i.a).As the reaction proceeded, its edges became smoother and later a transparent droplet layer formed, which had a newly formed solid phase embedded in it (Fig. (Fig. 2iv), and the surrounding aqueous layer consisted of Ca(NO 3 ) 2 (Fig. 2iii).
The particle morphology change shown in Fig. 2 was significantly different from the morphology change in the direct reaction of SO 2 with NO 2 (Zhao et al., 2018), where the CaCO 3 particle was first converted to a spherical Ca(NO 3 ) 2 droplet and then needle-shaped CaSO 4 crystals formed inside the droplet (Zhao et al., 2018).Moreover, the amount of CaSO 4 formed in this study was much higher than that in the direct reaction of SO 2 with NO 2 .The CaSO 4 solid particle constituted a significant fraction of the volume of the droplet, while in the direct reaction of SO 2 with NO 2 the few needleshaped CaSO 4 crystals that formed only constituted a small fraction of the droplet volume (Zhao et al., 2018).

Reaction process
During the reaction, the amounts of carbonate, nitrate, and sulfate were determined as a function of time, as shown in Fig. 3.At the beginning of the reaction, the amount of carbonate decreased slowly, while the amount of nitrate and sulfate increased slowly.After a period of induction of around 50 min, the reaction accelerated significantly, leading to a rapid consumption of carbonate and production of nitrate and sulfate.The decrease in the amount of carbonate and the increase in the amount of nitrate was because carbonate reacted continuously with NO 2 and H 2 O, forming Ca(NO 3 ) 2 .The detailed mechanisms of the multiphase reaction of carbonate with NO 2 and H 2 O were discussed in our previous studies (Li et al., 2010;Zhao et al., 2018).The mechanism of sulfate formation is discussed in detail in Sect.3.4 of the present study.Finally, the carbonate was completely consumed, and the amounts of nitrate and sulfate leveled off. Figure 3 shows that nitrate and sulfate were formed simultaneously during the reaction.This contrasts with the observations made during the direct reaction of SO 2 with NO 2 , where nitrate was formed first, and sulfate was essentially formed after the complete conversion of CaCO 3 particles to Ca(NO 3 ) 2 droplets (Zhao et al., 2018).Moreover, the time taken for carbonate to be completely consumed was longer in this study than in the direct reaction of SO 2 with NO 2 (∼ 120 vs. ∼ 40 min) when other conditions were kept the same (Zhao et al., 2018).

Reactive uptake coefficient of SO 2
The reactive uptake coefficients of SO 2 for sulfate formation (γ ) in the reaction of SO 2 with the O 2 / NO 2 / H 2 O / N 2 mixture on CaCO 3 with various O 2 concentrations are shown in Table 1.The value of γ for the reaction of SO 2 with O 2 / NO 2 at three O 2 concentrations (5, 20, and 86 %) was in the range of (0.35-1.7) × 10 −5 , and was 1.2 × 10 −5 in synthetic air.This latter value was 2-3 orders of magnitude higher than that for the reaction of SO 2 directly with NO 2 under similar conditions (Zhao et al., 2018).When other conditions were kept constant, γ increased with the O 2 concentration.This indicates that O 2 played a key role in enhancing the oxidation rate of SO 2 .
The role of O 2 in enhancing the reactive uptake of SO 2 reported here is consistent with the findings in some previous studies.For example, the data of Littlejohn et al. (1993) showed that the sulfite oxidation rate increases with the O 2 concentration (0-5 % by volume).Shen and Rochelle (1998) also found that in the presence of O 2 , the aqueous sulfite oxidation rate is enhanced.By investigating the oxidation of SO 2 by NO 2 in monodispersed water droplets growing on carbon nuclei, Santachiara et al. (1990) found that the sulfate formation rate with 2 % O 2 is much higher than that without O 2 .However, our findings, as well as those in the studies referred to above, are in contrast to those reported by Lee and Schwartz (1983), who found that changing from N 2 to air as a carrier gas only increases the bisulfite oxidation rate by no more than 10 %.The difference between our study and Lee and Schwartz (1983) could be due to the difference in O 2 diffusion from the gas to the condensed phase and the different mechanisms between the multiphase reaction on particles and the aqueous reaction.Only few studies have reported the S(IV) oxidation rate in the reaction of S(IV) with O 2 / NO 2 mixtures (Turšič et al., 2001;Littlejohn et al., 1993).However, due to the limiting step by the aqueous-phase mass transfer, it is difficult to quantitatively compare the reaction rates in those studies with the uptake coefficient in our study and the rate constants determined by Lee and Schwartz (1983) and Clifton et al. (1988).For example, a rate constant of 2.4 × 10 3 mol −1 L s −1 (at pH 3) can be derived from the results of Turšič et al. (2001), which is much lower than the values reported by Lee and Schwartz (1983) and Clifton et al. (1988).This can be attributed to the limiting step by the aqueous-phase mass transfer because the characteristic mixing time in the aqueous phase in Turšič et al. (2001) was likely much longer than that of Lee and Schwartz (1983) (1.7-5.3 s), according to the HSO − 3 concentration time series reported by Turšič et al. (2001).
It is important to note that the concentrations of NO 2 and SO 2 used in this study are much higher than those in the ambient atmosphere.High concentrations of reactant gases are often used in laboratory studies in order to simulate the ambient reactions at the timescale of days or weeks and to obtain high signal-to-noise ratios for detecting products within minutes or hours.In the ambient atmosphere, the reactive uptake coefficient of SO 2 should be lower than that in this study due to the lower NO 2 concentrations when other conditions are comparable, and the chemical and physical processes observed in this study, such as changes in particle composition, phase, hygroscopicity, and pH, should be much slower due to the lower concentrations of NO 2 and SO 2 .

Reaction mechanism
In the multiphase reaction of SO 2 with O 2 / NO 2 / H 2 O on CaCO 3 particles, we found that CaCO 3 reacted with NO 2 and H 2 O and produced Ca(NO 3 ) 2 , which deliquesced, forming liquid water, and provided a site for the aqueous oxidation of SO 2 .This process is similar to the direct reaction of SO 2 with NO 2 on CaCO 3 particles.The details of this part of the reaction mechanism were discussed in our previous study (Zhao et al., 2018).
Once the aqueous phase was formed, SO 2 could undergo multiphase reactions with O 2 / NO 2 .The mechanism of the direct aqueous reaction of S(IV) with NO 2 in the absence of www.atmos-chem-phys.net/18/6679/2018/Atmos.Chem.Phys., 18, 6679-6689, 2018 O 2 is complex.Previous studies have proposed two different mechanisms for the reaction.One involves SO q − 3 radical formation (Littlejohn et al., 1993;Shen and Rochelle, 1998;Turšič et al., 2001) (referred to as the "free-radical chain" mechanism), while the other involves the formation of NO 2 -S(IV) complexes (Clifton et al., 1988), but no radical formation (referred to as the "NO 2 -S(IV) complex" mechanism).
According to the NO 2 -S(IV) complex mechanism, the presence of O 2 should not affect the SO 2 oxidation rate; however, in this study, a substantial enhancement in the SO 2 oxidation rate was observed in the presence of O 2 compared with that in the absence of O 2 .Therefore, the NO 2 -S(IV) complex mechanism was less likely to have been important in this study.
In the free-radical chain mechanism, the SO q − 3 radical is proposed to be formed (Reaction R8, Table 2), which is based on the observation of S 2 O 2− 6 formation, with S 2 O 2− 6 , known to be the combination reaction product of SO q − 3 (Eriksen, 1974; Hayon et al., 1972;Deister and Warneck, 1990;Brandt et al., 1994;Waygood and McElroy, 1992).In addition to SO 2− 4 and NO − 2 , S 2 O 2− 6 was detected with an appreciable yield using Raman spectroscopy, following the reaction of NO 2 with aqueous sulfite (Littlejohn et al., 1993).S 2 O 2− 6 was also observed in the aqueous oxidation of bisulfite in an N 2 -saturated solution in the presence of Fe(III) using ion interaction chromatography (Podkrajšek et al., 2002).The SO q − 3 radical can react via two pathways, forming either S 2 O 2− 6 or SO 2− 4 (Reactions R9-R11, Table 2).Reactions R9-R11 have been well established in studies of S(IV) oxidation by other pathways, including OH oxidation, photooxidation, and transition metal catalyzed oxidation (Eriksen, 1974;Hayon et al., 1972;Deister and Warneck, 1990;Brandt et al., 1994;Brandt and Vaneldik, 1995;Waygood and McElroy, 1992).In addition, although previous studies have not reported the direct observation of the SO q − 3 radical in the aqueous reaction of S(IV) with NO 2 , SO q − 3 was observed in the reaction of NO − 2 with SO 2− 3 in an acidic buffer solution (pH = 4.0) using electron spin resonance (Shi, 1994).Because NO − 2 is formed in the aqueous reaction of SO 2 with NO 2 , and S 2 O 2− 6 as the combination reaction product of SO q − 3 is observed (Littlejohn et al., 1993), SO q − 3 formation is plausible.
In the presence of O 2 , the SO q − 3 radical can react rapidly with O 2 , forming the SO q − 5 radical (Reaction R12, Table 2).Following this reaction, a number of chain reactions can occur to ultimately form sulfate (Littlejohn et al., 1993;Seinfeld and Pandis, 2006;Shen and Rochelle, 1998) (Reactions R13-R16, Table 2).Littlejohn et al. (1993) observed that the amount of S 2 O 2− 6 relative to SO 2− 4 formed in the aqueous reaction of NO 2 with sulfite decreases in the presence of O 2 compared with the reaction in the absence of O 2 .At low NO 2 concentrations (< 5 ppm), S 2 O 2− 6 is undetectable in the presence of O 2 .This indicates that O 2 suppresses the reaction pathway of S 2 O 2− 6 formation (Reaction R9, Table 2).Because the SO q − 3 radical can react rapidly with O 2 , forming the SO q − 5 radical, and would therefore be consumed, the suppression of S 2 O 2− 6 formation can be attributed to the reaction of SO q − 3 with O 2 (Reaction R12, Table 2).Reactions (R12)-(R16) have been well established by studies of the oxidation of S(IV) by OH or photo-oxidation, and all the radicals have been observed (Hayon et al., 1972;Huie et al., 1989;Huie and Neta, 1987;Chameides and Davis, 1982;Seinfeld and Pandis, 2006).
The free-radical chain mechanism is consistent with the findings of this study and is therefore more plausible.The enhancement of the SO 2 oxidation rate in the reaction of SO 2 with O 2 / NO 2 / H 2 O on CaCO 3 particles compared with that in the direct reaction of SO 2 with NO 2 / H 2 O was attributed to O 2 .Although during the reaction in the absence of O 2i.e., the direct oxidation of SO 2 by NO 2 -the SO q − 3 radical can be formed (Reaction R7), the reaction chain cannot propagate (Reactions R12-R16).Therefore, the S(IV) oxidation rate and the reactive uptake coefficient of SO 2 were much lower than those in the presence of O 2 .According to the difference between the reactive uptake coefficient in this study and in the direct reaction of SO 2 with NO 2 (Zhao et al., 2018), the sulfate production rate via chain reactions due to the presence of O 2 (20 %) was 2-3 orders of magnitude faster than the direct oxidation of SO 2 by NO 2 .This indicates that sulfate production in the reaction of SO 2 with O 2 / NO 2 was largely due to O 2 oxidation via the chain reaction pathway, i.e., "autoxidation" of S(IV), rather than the direct oxidation of SO 2 by NO 2 , and thus O 2 was the main oxidant of SO 2 .
In addition to the two mechanisms above, Spindler et al. (2003) proposed a reaction mechanism involving first q − 3 concentration, shifting the equilibrium to the product side and promoting the overall S(IV) oxidation.O 2 can act in a similar way as in the freeradical chain mechanism.Admittedly, we cannot rule out the possibility of a NO 2 -S(IV) complex formation.But such a mechanism can still be classified as the free-radical chain mechanism since the S(IV) oxidation still proceeds via the radical chain reactions.Although the direct oxidation of SO 2 by NO 2 only accounted for a very small fraction of sulfate formation, NO 2 played an important role in the SO 2 oxidation by initiating the chain reactions via the production of the SO q − 3 radical (Reaction R7).In the experiment without NO 2 , but with other reaction conditions the same, we were unable to detect sulfate after 5 h of reaction.This indicates that O 2 by itself cannot initiate the chain reactions (although it favors chain propagation), and that the oxidation of SO 2 by O 2 was slow.The effect on the SO 2 oxidation rate when both NO 2 and O 2 were present was much higher than the sum of the effect of NO 2 and O 2 .We refer to this effect as the synergy of NO 2 and O 2 , which resulted in the fast oxidation of SO 2 to form sulfate in this study.This effect is similar to a "ternary" reaction found with the reaction of NO 2 -particles-H 2 O or SO 2 -particles-O 3 (Zhu et al., 2011), where the reaction rate can be much faster than the sum of the reaction rates for the reaction of the second and third reactant with the first reactant.In addition to acting as the initiator of chain reactions, NO 2 also contributed to the formation of the aqueous phase through the reaction with CaCO 3 , forming Ca(NO 3 ) 2 as discussed above, which provided a site for S(IV) oxidation.
Based on the discussion above, we summarize the reaction mechanism that occurred in this study in Table 2.The reac-tions are classified as chain initiation, chain propagation, and chain termination.The dominant S(IV) species depends on pH.Due to the fast dissociations of SO 2 q H 2 O and HSO − 3 , reactions consuming one of these S(IV) species will result in instantaneous re-establishment of the equilibria between them (Seinfeld and Pandis, 2006).In this study, the pH of the aqueous layer of Ca(NO 3 ) 2 may change dynamically with time during the reaction and may not be completely homogeneous within the aqueous droplet.The pH values could vary between ∼ 3 and ∼ 7.6.In the surface of the aqueous layer, pH was mainly determined by the gas-aqueous equilibrium of SO 2 , and was estimated to be ∼ 3.In the vicinity of the CaCO 3 core, pH was mainly determined by the hydrolysis of carbonate and was estimated to be ∼ 7.6.It is likely that both HSO − 3 and SO 2− 3 were present, and the dominant species depended on the reaction time and location within the aqueous droplet.Nevertheless, to make the reaction mechanism clearer, HSO − 3 was used in the reaction equations.Similar reaction equations are also applicable to SO 2− 3 because of the fast dissociations of SO 2 q H 2 O and HSO − 3 .Overall, the reaction can be written as follows, which clearly shows that O 2 was the main oxidant for sulfate formation: this.First, the CaSO 4 q 2H 2 O formed in the reaction could cover the CaCO 3 surface and partly suppress the diffusion of aqueous ions, such as protons, and also limit the contact of reactants with the surface of the CaCO 3 particles, thus reducing the CaCO 3 consumption rate.Second, compared with the direct reaction of SO 2 with NO 2 , a much higher fraction of CaCO 3 was converted to CaSO 4 q 2H 2 O instead of Ca(NO 3 ) 2 due to the fast production of CaSO 4 q 2H 2 O. Therefore, the volume of a Ca(NO 3 ) 2 droplet was much smaller than that in the direct reaction of SO 2 with NO 2 for a given CaCO 3 particle.Because the uptake rate of NO 2 was proportional to the droplet surface area and the NO 2 hydrolysis rate was proportional to the droplet volume, the rate of nitric acid production from NO 2 hydrolysis and its reaction rate with CaCO 3 were reduced.Therefore, the CaCO 3 particles were consumed more slowly in the reaction with O 2 / NO 2 .

Conclusion and implications
We investigated the multiphase reaction of SO Using a method developed in our previous study (Zhao et al., 2018), we assessed the importance of the multiphase oxidation of SO 2 by O 2 in the presence of NO 2 by estimating the lifetime of SO 2 due to multiphase reactions and the lifetime due to the gas-phase reaction (with the OH radical).The lifetime of SO 2 due to the multiphase reaction of SO 2 with O 2 / NO 2 was estimated to be around 20 days using the reactive uptake coefficient of SO 2 (1.2 × 10 −5 ) and the typical particle surface area concentration for mineral aerosols in winter in Beijing (6.3 × 10 −6 cm 2 cm −3 ) (Huang et al., 2015).This lifetime is comparable to the lifetime of SO 2 due to the gas-phase reaction with OH, which is ∼ 12 days, assuming that the daytime OH concentration is 1 × 10 6 molecules cm −3 (Lelieveld et al., 2016;Prinn et al., 2005).Therefore, we conclude that the multiphase oxidation of SO 2 by O 2 in the presence of NO 2 is likely to be an important source of sulfate and a sink of SO 2 in the ambient atmosphere, and can play a significant role in the sulfate formation during severe haze episodes, such as those that frequently occur in China.During haze episodes, there are high concentrations of SO 2 and NO 2 and relative humidity is often high (Zhang et al., 2014;Zheng et al., 2015b;Wang et al., 2016).Under these conditions, the multiphase oxidation of SO 2 by O 2 in the presence of NO 2 could proceed rapidly, forming sulfate.The enhanced sulfate concentration due to multiphase reactions and resulting aerosol water content can further promote the multiphase oxidation of SO 2 .The reaction thus proceeds in a self-accelerating way.Therefore, it can contribute significantly to sulfate formation during haze episodes, which could explain the discrepancies between the observed and modeled sulfate concentrations (Cheng et al., 2016;Gao et al., 2016;Wang et al., 2016;Zheng et al., 2015a).
In addition, elucidating the mechanism of the multiphase reaction of SO 2 with O 2 / NO 2 / H 2 O in the atmosphere is important for the other atmospheric implications of the reaction besides sulfate formation.According to the reaction mechanism, the direct oxidation of SO 2 by NO 2 forms sulfate and nitrite, with a stoichiometry of 1 : 1, and nitrite can further form HONO under acidic conditions.The HONO could then evaporate into the atmosphere, where it would be an important source of OH radicals.If NO 2 were the main oxidant of SO 2 in the multiphase reaction, the reaction would form one HONO molecule for every sulfate molecule formed.Thus, the oxidation of SO 2 by NO 2 can simultaneously be an important source of HONO and OH radicals, and SO 2 oxidation would be strongly coupled with reactive nitrogen chemistry.However, according to the mechanism of this study, NO 2 only acted as an initiator of the chain reactions in SO 2 oxidation and essentially all the SO 2 was oxidized by O 2 .Therefore, the amount of HONO formation per sulfate molecule formed was trivial.The oxidation of SO 2 by O 2 / NO 2 is expected to be neither an important source of HONO and OH in the atmosphere nor to have a significant influence on reactive nitrogen chemistry.
In this study, we investigated the reaction of SO 2 with O 2 in the presence of NO 2 at three O 2 concentrations.The influence of the O 2 concentration was shown to be significant.Future experimental results with more O 2 concentration levels would provide more insight into the reaction mechanism and process.
In addition, in the ambient atmosphere, the internal mixing of organics with S(IV) in particles may influence the S(IV) oxidation rate by O 2 in the presence of NO 2 .When organics are abundant in particles, for example during haze episodes in China, they can react with and thus scavenge radical anion carriers such as SO q − 5 and SO q − 4 (Herrmann, 2003;Herrmann et al., 2015;Huie, 1995).Therefore, the presence of internally mixed organics can reduce the effectivity of the po-Atmos.Chem.Phys., 18, 6679-6689, 2018 www.atmos-chem-phys.net/18/6679/2018/tential radical reaction chain and of S(IV) oxidation, which can undermine the importance of the oxidation by O 2 in the presence of NO 2 in the overall S(IV) oxidation.
Data availability.The data in the figures in both the main text and the Supplement are available upon request to the corresponding author (tzhu@pku.edu.cn).
Competing interests.The authors declare that they have no conflict of interest.

Figure 1 .
Figure1shows the Raman spectra of a CaCO 3 particle during the multiphase reaction of SO 2 with O 2 / NO 2 / H 2 O on its surface.The peak at 1087 cm −1 was assigned to the symmetric stretching of carbonate (ν s (CO 2− 3 ))(Nakamoto, 1997).During the reaction, the peak at 1087 cm −1 decreased continuously and finally disappeared as new peaks were observed.The peak at 1050 cm −1 was assigned to the symmetric stretching of nitrate (ν s (NO − 3 )).The peaks at 1010 and 1136 cm −1 were assigned to the symmetric stretching (ν s (SO 2− 4 )) and asymmetric stretching (ν as (SO 2− 4 )) of sulfate in gypsum (CaSO 4 q 2H 2 O), respectively(Sarma et al., 1998).In addition, after the reaction, a broad envelope in the range of 2800-3800 cm −1 assigned to the stretching of the OH bond in water molecules was observed.Above this envelope, there were two peaks at 3408 and 3497 cm −1 , which were assigned to OH bond stretching in crystallization water of CaSO 4 q 2H 2 O(Sarma et al., 1998;Ma et al., 2013).During the multiphase reaction with the SO 2 / O 2 / NO 2 / H 2 O mixture, the CaCO 3 particles displayed a remarkable change in morphology.The original CaCO 3 particle was a rhombohedron crystal (Fig.2i.a).As the reaction proceeded, its edges became smoother and later a transparent droplet layer formed, which had a newly formed solid phase embedded in it (Fig.2i.d).The size of the new solid phase grew during the reaction (Fig.2i.d-f) and it seemed to contain many micro-crystals.Raman mapping revealed that the new solid phase consisted of CaSO 4 q 2H 2 O

Figure 2 .
Figure 2. Microscopic image (i) and Raman mapping image of carbonate (ii), nitrate (iii), and sulfate (iv) on the CaCO 3 particle during the multiphase reaction SO 2 with O 2 / NO 2 / H 2 O on the particle.Panels (a-f) correspond to the reaction times of 0, 20, 41, 76, 117, and 193 min.SO 2 : 75 ppm, NO 2 : 75 ppm, RH: 72 %, O 2 : 20 %.The mapping images of carbonate, nitrate, and sulfate are made using the peak area at 1050, 1010, and 1087 cm −1 , respectively.The red, blue, and green colors indicate the peak intensity of carbonate, nitrate, and sulfate, respectively.The dashed lines in panel (iii.f)and (iv.f) indicate the shape of the droplet at the end of the reaction.

Table 1 .
Reactive uptake coefficient of SO 2 for sulfate formation at 82 % RH and at different O 2 concentrations.SO 2 / NO 2 / O 2 concentration γ 2 with O 2 / NO 2 / H 2 O on CaCO 3 particles.The reaction converted CaCO 3 particles to Ca(NO 3 ) 2 droplets, in which CaSO 4 q 2H 2 O was embedded and accounted for a significant fraction of the droplet volume by the end of the reaction.The Ca(NO 3 ) 2 droplet formed by the reaction of CaCO 3 with NO 2 provided a site for the multiphase oxidation of SO 2 .Generally, nitrate and sulfate were formed simultaneously.The reactive uptake coefficient of SO 2 for sulfate formation in the reaction of SO 2 with NO 2 / H 2 O in synthetic air was determined to be around 10 −5 .Compared with the reaction of SO 2 with NO 2 on a CaCO 3 particle in the absence of O 2 , i.e., the direct oxidation of SO 2 by NO 2 in N 2 , the sulfate production rate in the reaction of SO 2 with O 2 / NO 2 was enhanced by 2-3 orders of magnitude.According to the findings of this study and the existing literature, SO 2 oxidation likely proceeded via a free-radical chain reaction mechanism.O 2 was the main oxidant of SO 2 , and NO 2 mainly acted as an initiator of the chain reactions.The synergy of NO 2 and O 2 resulted in the fast oxidation of SO 2 .The absence of either NO 2 or O 2 led to much slower SO 2 oxidation.