Enhanced heterogeneous uptake of sulfur dioxide on mineral particles through modification of iron speciation during simulated cloud processing

Abstract. Iron-containing mineral aerosols play a key role in the oxidation
of sulfur species in the atmosphere. Simulated cloud processing (CP) of
typical mineral particles, such as illite (IMt-2), nontronite (NAu-2),
smectite (SWy-2) and Arizona Test Dust (ATD) is shown here to modify sulfur
dioxide (SO2) uptake onto mineral surfaces. Heterogeneous oxidation of
SO2 on particle surfaces was firstly investigated using an in situ DRIFTS
apparatus (diffuse reflectance infrared Fourier transform spectroscopy). Our results showed that the Brunauer–Emmett–Teller (BET) surface area normalized uptake
coefficients (γBET) of SO2 on the IMt-2, NAu-2, SWy-2 and
ATD samples after CP were 2.2, 4.1, 1.5 and 1.4 times higher than the
corresponding ones before CP, respectively. The DRIFTS results suggested
that CP increased the amounts of reactive sites (e.g., surface OH groups) on
the particle surfaces and thus enhanced the uptake of SO2. Transmission
electron microscopy (TEM) showed
that the particles broke up into smaller pieces after CP, and thus produced
more active sites. The “free-Fe” measurements confirmed that more reactive
Fe species were present after CP, which could enhance the SO2 uptake
more effectively. Mössbauer spectroscopy further revealed
that the formed Fe phases were amorphous Fe(III) and nanosized ferrihydrite
hybridized with Al ∕ Si, which were possibly transformed from the Fe in the
aluminosilicate lattice. The modification of Fe speciation was driven by the
pH-dependent fluctuation coupling with Fe dissolution–precipitation
cycles repeatedly during the experiment. Considering both the enhanced SO2
uptake and subsequent promotion of iron dissolution along with more active
Fe formation, which in turn led to more SO2 uptake, it was proposed
that there may be a positive feedback between SO2 uptake and iron
mobilized on particle surfaces during CP, thereby affecting climate and
biogeochemical cycles. This self-amplifying mechanism generated on the
particle surfaces may also serve as the basis of high sulfate loading in
severe fog–haze events observed recently in China.


Abstract. Iron-containing mineral aerosols play a key role in the oxidation of sulfur species in the atmosphere. Simulated cloud processing (CP) of typical mineral particles, such as illite (IMt-2), nontronite (NAu-2), smectite (SWy-2) and Arizona Test Dust (ATD) is shown here to modify sulfur dioxide (SO 2 ) uptake onto mineral surfaces. Heterogeneous oxidation of SO 2 on particle surfaces was firstly investigated using an in situ DRIFTS apparatus (diffuse reflectance infrared Fourier transform spectroscopy). Our results showed that the Brunauer-Emmett-Teller (BET) surface area normalized uptake coefficients (γ BET ) of SO 2 on the IMt-2, NAu-2, SWy-2 and ATD samples after CP were 2.2, 4.1, 1.5 and 1.4 times higher than the corresponding ones before CP, respectively. The DRIFTS results suggested that CP increased the amounts of reactive sites (e.g., surface OH groups) on the particle surfaces and thus enhanced the uptake of SO 2 . Transmission electron microscopy (TEM) showed that the particles broke up into smaller pieces after CP, and thus produced more active sites. The "free-Fe" measurements confirmed that more reactive Fe species were present after CP, which could enhance the SO 2 uptake more effectively. Mössbauer spectroscopy further revealed that the formed Fe phases were amorphous Fe(III) and nanosized fer-rihydrite hybridized with Al / Si, which were possibly transformed from the Fe in the aluminosilicate lattice. The modification of Fe speciation was driven by the pH-dependent fluctuation coupling with Fe dissolution-precipitation cycles repeatedly during the experiment. Considering both the enhanced SO 2 uptake and subsequent promotion of iron dissolution along with more active Fe formation, which in turn led to more SO 2 uptake, it was proposed that there may be a positive feedback between SO 2 uptake and iron mobilized on particle surfaces during CP, thereby affecting climate and biogeochemical cycles. This self-amplifying mechanism generated on the particle surfaces may also serve as the basis of high sulfate loading in severe fog-haze events observed recently in China.
consists of quartz, feldspars, carbonates (calcite, dolomite) and clay minerals (illite, kaolinite, chlorite, montmorillonite), the exact composition varies with source (Claquin et al., 1999;Formenti et al., 2008;Journet et al., 2008). A long-range transport would result in a decrease in quartz relative to the clay fraction because of the more rapid removal of quartz, hence clay is an important component of mineral dusts Journet et al., 2008). During the long-range transport, mineral dust provides a reactive surface for heterogeneous chemistry (Zhang et al., 2006;George et al., 2015;Huang et al., 2015). Heterogeneous reactions of atmospheric trace gases on mineral dust particles are of great significance as these reactions alter the chemical balance of the atmosphere and modify the properties of individual particles (Usher et al., 2003;Wu et al., 2011;Huang et al., 2015).
SO 2 is an important trace gas, which is released mainly by fossil fuel combustion and volcanic emission. The heterogeneous conversion of SO 2 on mineral dust surfaces leads to the formation of sulfuric acid and sulfate aerosols, resulting in a significant cooling effect on the global climate by scattering solar radiation and acting as cloud condensation nuclei (CCN) to affect climate indirectly (Lelieveld and Heintzenberg, 1992;Usher et al., 2003;Kolb et al., 2010). In addition, sulfate-containing particles play a significant role in the haze formation in China in recent years Yang et al., 2017). SO 2 can be oxidized to sulfate by OH radical in the gas phase, be oxidation in cloud and fog droplets by ozone and hydrogen peroxide in the aqueous phase (Luria and Sievering, 1991), or through heterogeneous processes that occur on aerosol particle surfaces (Usher et al., 2003;Ullerstam et al., 2003). However, the high sulfate levels measured in recent field observations cannot be explained by current atmospheric models (Kerminen et al., 2000;Wang et al., 2003;Cheng et al., 2016), leading to a large gap between the modeled and field-observed sulfate concentrations using known oxidation pathways (Herman, 1991;Kasibhatla et al., 1997;Barrie et al., 2016). Overall, on a global scale, atmospheric SO 2 concentrations were typically overestimated, while sulfate tended to be underestimated, suggesting missing sulfate production pathways (Harris et al., 2013;Kong et al., 2014).
It has been suggested that the heterogeneous conversion of SO 2 could make an important contribution to the atmospheric sulfate loading. Laboratory studies typically focus on SO 2 uptake onto a variety of metal oxides and mineral particles (Goodman et al., 2001;Usher et al., 2002;Zhao et al., 2015;Yang et al., 2016) and have confirmed that its conversion rate on the surface of Fe (hydr)oxides was faster compared to other metal oxides investigated, in good agreement with the field measurements (Usher et al., 2002;Zhang et al., 2006). Atmospheric Fe is emitted from both anthropogenic (primarily biomass burning, coal and oil combustion) and natural (mineral dust and volcanic ash) sources, with the mineral dust source dominant globally (Siefert et al., 1998;Luo et al., 2008;Ito et al., 2019). It has been established that an important in-cloud S(IV) oxidation pathway is catalyzed by natural transition-metal ions, especially Fe hosted within mineral particles (Alexander et al., 2009;Harris et al., 2013).
Another important consideration for heterogeneous chemistry of mineral dust aerosol is how mineral dust particles change in the atmosphere. During long-range transport, mineral particles often undergo chemical ageing by atmospheric processes Baker and Croot, 2010;. Cloud processing involves cloud water condensation and evaporation on the particle surfaces, along with drastic liquid water content and pH fluctuations (Mackie, 2005;Rubasinghege et al., 2016). During cloud processing (CP), the high relative humidity (RH) results in high aerosol water content and relatively high pH (Behra et al., 1989;Baker and Croot, 2010;. When water evaporates from cloud droplets to wet aerosol at higher temperature, the particles only contain a concentrated aqueous aerosol solution in which the pH can be lower than 2 (Zhu et al., 1993;Meskhidze, 2003;Shi et al., 2015). Therefore, there is a highly acidic film (e.g., pH = 2) in the "wet aerosol" phase versus a less acidic droplet (near neutral, 5-6) in the "cloud droplet" phase within clouds (Shi et al., 2015). During its lifetime, a typical aerosol particle may experience several cloud cycles involving large pH variations before being removed from the atmosphere as rain or through dry deposition (Pruppacher and Jaenicke, 1995;Maters et al., 2016). Herein, the simulated CP experiment was conducted by changing pH between 2 and 5-6, in accordance with the previous studies (Spokes et al., 1994;Mackie, 2005;Shi et al., 2009).
It has been well documented that pH is especially important for Fe mobilization (Zhu et al., 1993;Desboeufs et al., 2001;Deguillaume et al., 2010;Maters et al., 2016). The fluctuating pH during CP will impact and change the Fe speciation and morphology in dust particles (Zhuang et al., 1992;Wurzler et al., 2000;Shi et al., 2009;Kadar et al., 2014). The low pH will increase Fe solubility and bioavailability of dust during transport, thereby providing Fe external input to the open ocean surface to promote marine primary productivity (Spokes et al., 1994;Desboeufs et al., 2001). It has been found that Fe-rich nanoparticle aggregates were formed from Saharan soil and goethite upon simulated CP conditions, in good agreement with their field measurements from the wet-deposited Saharan dusts collected from the western Mediterranean (Shi et al., 2009). Fe nanoparticles are more chemically reactive (Wurzler et al., 2000;Desboeufs et al., 2001), and possibly lead to a remarkable difference in heterogeneous chemistry. However, little is known about the influence of CP on SO 2 uptake onto particle surfaces up to now.
In this study, we employed four typical Fe-containing mineral samples as surrogates to perform simulated CP experiments. The SO 2 uptakes on the mineral particles before and after CP were compared using in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). Transmis-Atmos. Chem. Phys., 19, 12569-12585, 2019 www.atmos-chem-phys.net/19/12569/2019/ sion electron microscopy (TEM) was applied to observe the morphological and mineralogical change of mineral particles. The Fe speciation modification during simulated CP was further monitored by the dissolved Fe measurement, the free-Fe analysis and Mössbauer spectroscopic characterization.
2 Materials and methods

Mineral particles
The standard mineral samples of IMt-2, NAu-2 and SWy-2 were purchased from the Source Clay Minerals Repository (Purdue University, West Lafayette, IN, USA). ATD was purchased from Powder Technology Inc. (Burnsville, MN, USA). The mineral samples were coarsely ground using a mortar and pestle before being more finely ground using an all-dimensional planetary ball mill QM-QX (Nanjing University Instrument Plant) and were sieved to particle diameters (D p ) < 45 µm prior to analysis. The Brunauer-Emmett-Teller (BET) specific surface areas (S BET ) of the samples were measured with a Quantachrome Nova 1200 BET apparatus. Total iron content (Fe T ) of the samples was determined using inductively coupled plasma atomic emission spectroscopy (ICP-AES, Jobin Yvon Ultima). The chemical compositions of the particles were analyzed by X-ray fluorescence spectrometry (XRF, PANalytical Axios Advanced).

Cloud processing simulation experiment
The simulated CP experiments were conducted at a constant temperature (298±1 K) using a Pyrex glass vessel with a water jacket. The suspensions contained a mineral particle loading of 1 g L −1 and were subjected to acidic (pH = 2 ± 0.1, 24 h) and near-neutral pH (pH = 5-6, 24 h) cycles for one to three times according to the previous methods (Spokes et al., 1994;Mackie, 2005;Shi et al., 2009). Suspension pH was adjusted by adding dilute H 2 SO 4 or NH 4 OH. The CaCO 3 equivalent alkalinity of the dust was determined in accordance with APHA (American Public Health Association) method 2320B so that acid additions to control pH could be adjusted accordingly (Mackie, 2005). The amount of acid or alkali added to achieve these pH cycles was less than 1 % of the total volume of the suspensions. The experiments were performed under a constant stirring (about 50 rpm) in the dark for 144 h. At the end of the CP experiment, the suspensions were filtered through 0.2 µm PTFE filters (Millipore). The filter residue was air-dried and was further applied to the DRIFTS experiment, as well as TEM observation, free-Fe measurement and Mössbauer spectroscopic characterization.

SO 2 uptake on the mineral particles
The SO 2 uptake on the particle surfaces before and after CP was investigated by a Shimadzu Tracer-100 FTIR spectrometer equipped with a high-sensitivity mercury cadmium telluride (MCT) detector and a diffuse reflectance accessory. A temperature controller was fitted to the DRIFTS chamber to ensure constant reaction temperature (298 K). Weighted samples were placed into a ceramic crucible (0.35 mm depth, 5 mm i.d.) in the chamber. Mass flow controllers (Beijing Sevenstar electronics Co., LTD) were used to adjust the reactant gases to a flux with expected concentration and relative humidity. The sample was firstly pretreated in a 100 mL min −1 flow of synthetic air (21 % O 2 and 79 % N 2 ) for 1 h to blow off water and impurities on particle surfaces. When the background spectrum of the fresh sample reached steady state, the reactant gas of SO 2 (5.0 ppm) along with synthetic air was introduced into the chamber at a total flow rate of 120 mL min −1 for 45 min, during which the IR spectrum was recorded automatically every 5 min at a resolution of 4 cm −1 for 100 scans in the spectral range of 900 to 4000 cm −1 . Atmospheric moisture was simulated with a RH level around 40 % by guiding one high-purity air flux through water. The humidity value was monitored using a hygrometer.
The sulfate products were analyzed by ion chromatograph (IC) after the DRIFTS experiments. The particles were extracted with 5 mL ultrapure water by an ultrasonic extractor. After 10 min, the extracted solution was passed through a 0.22 µm PTFE membrane filter and the leaching solution was analyzed using a Metrohm 883 Basic IC equipped with an A5-250 column. A weak base eluent (3.2 mmol L −1 Na 2 CO 3 plus 1.0 mmol L −1 NaHCO 3 ) was used for anion detection at a flow rate of 0.70 mL min −1 . To discriminate the adsorbed sulfate during simulated CP experiment and the sulfate ions generated from the heterogeneous reaction, the adsorbed sulfate on the particles during simulated CP experiment was initially measured as blank. The heterogeneous uptake of SO 2 was calculated by subtracting the blank value from the total sulfate ions.
The reactive uptake coefficient (γ ) was defined as the rate of sulfate formation on the surface (d SO 2− 4 /dt, ions per second) divided by collision frequency (Z, molecules per second) (Usher et al., 2003;Ullerstam et al., 2003;Kong et al., 2014;Huang et al., 2015). where A s is the effective sample surface of the samples (m 2 ), ν is the mean molecular velocity of SO 2 (m s −1 ), R is the gas constant (J mol K −1 ), T is the absolute temperature (K), and M SO 2 is the molecular weight of SO 2 (kg mol −1 ). A conversion factor was obtained by a calibration plot with numbers of SO 2− 4 analyzed by ion chromatography (IC, Metrohm 883 Basic, Switzerland) versus the integrated areas of sulfate products from DRIFTS spectra. The residual sulfate during simulated CP experiments were deducted as background. The calculated conversion factor of SO 2− 4 is 1.170 × 10 15 (ions per integrated units). Integrated areas for the total sulfur-containing products were calculated to show the maximal sulfate formation rates. The reactive uptake coefficients for SO 2 were determined to be γ BET and γ geo using the BET area (A BET = mass × S BET ) and geometric area (A geo = mass × S geo ) as the reactive area, respectively.

Morphological and mineralogical characterization of the Fe speciation
A FEI TECNAI G2 S-TWIN F20 TEM equipped with an Oxford energy-dispersive X-ray (EDX) spectrometer was used to analyze the morphological and chemical composition of individual particles before and after CP. Suspensions (0.2 g L −1 ) of each particle were prepared in methanol and sonicated for at least 1 h. A drop of this suspension was then applied to a carbon-coated Cu TEM grid (400 mesh; EMS) and allowed to air-dry. The operation was conducted in bright field mode at 120 kV. The Fe contents of the typical individual mineral particle were calculated from the values of 50 typical particles. To obviously observe the morphological changes, high-resolution TEM (HRTEM) images were also collected to observe nanoscale structural features, e.g., surface roughness and lattice fringes. The content of free Fe in the mineral particles was determined by a citrate-buffered-dithionite (CBD) sequential Fe extractions method according to the literature (Lafon et al., 2004;Shi et al., 2009). Simply, 30 mg of the dust samples was treated for 24 h with a 10 mL ascorbate solution (pH = 7.5) to extract chemically highly labile Fe phases (Fe A ), mainly composed of amorphous, nanoparticle and/or poorly crystalline ferrihydrite. The solutions were filtered through 0.2 µm polycarbonate filters. The dust particles collected on the filters were subsequently treated for 2 h with a 10 mL sodium dithionite solution (pH = 4.8) to extract crystalline Fe (oxyhydr)oxides (Fe D ), which are mainly goethite and hematite. After each reaction step, the dissolved Fe concentrations (Fe A and Fe D ) in the filtrates were determined using ICP-AES. The sum of these two pools (Fe A + Fe D ) was defined as the "free-Fe" fraction . The other fraction was denoted as the "structural-Fe" fraction in aluminosilicate crystals, which could be calculated from the difference between the Fe T and free-Fe fractions (Lafon et al., 2004).
The Mössbauer spectroscopic analysis performed in transmission geometry with a constant acceleration was used to inspect the chemical valence and the surrounding structure of Fe in the particles before and after CP. 57 Co was used as the Mössbauer source, and a 1 mm thick Na(TI) scintillator coupled to a EMI9750B photoelectric multiplier was used as the detector (Cwiertny et al., 2008). The measurement was carried out at room temperature (RT) with a duration of 24 h for one sample (around 1.5×10 6 counts per channel). Experimental data were fitted by a least-squares fitting program. The isomer shift values were calibrated against a spectrum for α-Fe metal foil.
During the simulated CP experiment, the total dissolved iron (Fe s ) and the dissolved Fe(II) in the suspensions were measured colorimetrically by the ferrozine method, as described in previous studies (Viollier et al., 2000;Cwiertny et al., 2008). For Fe(II) analysis, 200 mL of a 5 mM 1,10phenanthroline solution and 200 mL of an ammonium acetate buffer were added into 1 mL of sample. To avoid possible interference from Fe(III), which can also form a complex with 1,10-phenanthroline when present at high concentrations, 50 mL of 0.43 M ammonium fluoride was added to the sample prior to 1,10-phenanthroline. The mixture was allowed to sit in the dark for 30 min prior to ultravioletvisible spectroscopy (UV-Vis) analysis, during which time a reddish-orange color developed if Fe(II) was present. Fe s was determined via the same protocol, except that 20 mL of 1.5 M hydroquinone, which reduces Fe(III) to Fe(II), was added to the sample rather than ammonium fluoride. Absorbance measured at 510 nm was converted to concentrations using aqueous standards prepared from anhydrous beads of ferrous chloride. Standards were prepared in each acid used in dissolution studies, and no matrix effects were observed. These conditions resulted in a detection limit of 1 µM. The concentration of dissolved Fe(III) was calculated from the difference in experimentally measured concentrations of total dissolved iron and dissolved Fe(II).
Additionally, the dissolved Fe(III) could precipitate out as the pH increased, and then the Fe mineralogy of the deposit was also observed. NAu-2 released about 300 µM of dissolved Fe at pH 2. The dissolving solution (200 mL) was sampled after filtration (0.2 µm polycarbonate filter). The clear solution was subjected to changing acidity from pH 2 to 5 by the stepwise addition of dilute NH 4 OH. The precipitated particles were separated out by 0.2 µm filters and were used in TEM and Mössbauer analysis. Size distributions for the Fe-bearing particles formed in the suspensions were determined by a Horiba LB-500 light scattering microscope within the size range of 3-6000 nm.  Figure 1 shows the TEM images of the mineral particles before and after CP. As shown in Fig. 1a, c, e and g, the IMt-2, NAu-2, SWy-2 and ATD samples before CP primarily consisted of laminar aluminosilicate with irregular shape and rough morphologies mainly at the micrometer scale, all of which were characterized by various fractions of Fe (1.5 %-26.2 %), along with minor Mg (0.1 %-16.5 %), K (0.0 %-7.8 %) and Ca (0.0 %-1.1 %). The Fe within the aluminosilicates of the particles was evenly distributed. Besides, some Fe-rich crystals of several hundreds of nanometers in size were found to attach onto the ATD particles, which were identified as α-Fe 2 O 3 from the typical d spacing analysis of HRTEM (Janney et al., 2000).
After the simulated CP, all of the processed mineral particles showed much smaller size than the ones before CP. For example, the typical IMt-2 and NAu-2 particles after CP ( Fig. 1b and d) were < 1 µm in size. Under the TEM, the average Fe content of the individual IMt-2 and SWy-2 particles ( Fig. 1b and f) decreased from 5.5 % (±1.9 %; n = 50) to 4.1 % (±1.6 %; n = 50) and from 2.4 % (±0.6 %; n = 50) to 2.1 % (±0.5 %; n = 50), respectively. In addition, the IMt-2 particles after CP showed a heterogeneous distribution of the Fe on the basis of the EDX data. Most of the aluminosilicate in IMt-2 after CP hosted lower Fe content (4.1 %), whereas a few of the Fe-rich particles with less Si / Al were observed with irregular shapes at the nanoscale. The TEM images of the NAu-2 and ATD particles after CP (Fig. 1h) showed some pseudohexagonal nanoparticles with around 5 nm in diameter. Based on the EDX and selected area electron diffraction (SAED) analysis, these nanoparticles were Fe-rich and the d spacings was at about 1.5-2.5 Å, all of which were identified to be 2-line ferrihydrite (Janney et al., 2000;Shi et al., 2009).
The TEM observation suggested that CP induced the disintegration of mineral particles and thus produced enhanced surface area, resulting in more active sites available on the particle surfaces for SO 2 uptake. Results of TEM also showed that CP influenced the Fe mineralogy and lead to the Fe-rich nanoparticle formation, which could partly explain the higher SO 2 uptake on the mineral particles after CP.

Effect of simulated CP on heterogeneous transformation of SO 2
The in situ DRIFTS spectra on the IMt-2, NAu-2, SWy-2 and ATD samples before and after CP exposed to SO 2 as a function of time are shown in Fig. 2. For the IMt-2 sample before CP ( Fig. 2a and b), the intensities of the broad peaks from 3600 to 3000 cm −1 and a weak peak at 1650 cm −1 increased with time. The band between 3600 and 3000 cm −1 was attributed to the vibrations of hydrogenbonded hydroxyl species , while the absorption peak at 1650 cm −1 was mainly associated with H 2 O produced from the reaction between SO 2 and surface hydroxyls (Nanayakkara et al., 2012;Cheng et al., 2016). A weak vibration was observed at around 1100 cm −1 , which might be attributed to free sulfate anions on the particle surface (Ullerstam et al., 2003;Nanayakkara et al., 2012;Yang et al., 2016). Previous studies established that various types of surface OH groups are the key reactive sites for sulfite or sulfate and bisulfite or bisulfate formation on mineral oxides (Faust et al., 1989;Usher et al., 2003;Ullerstam et al., 2003), because of the complexes formed between sulfite or sulfate species and the surface OH. Generally, the SO 2 adsorption grows in intensity with decreasing OH stretching and H 2 O banding (Zhang et al., 2006). However, the OH peaks herein were not observed to decrease with prolonged time because the losses of H 2 O and OH groups on the particle surfaces were replenished by maintaining the constant RH in this study. When the same set of experiments was carried out using the IMt-2 sample after CP (Fig. 2b), the intensities of the prominent peaks were significantly higher than those for the IMt-2 sample before CP. Four new bands were readily observed at 1167, 1100, 1088 and 1077 cm −1 . The new bands were easily assigned to the stretching motion of surfacecoordinated sulfate species (1167 cm −1 ), i.e., bidentate surface sulfate complexes, free sulfate ion (1100 cm −1 ), and sulfite or bisulfite species (1088 and 1077 cm −1 ) (Peak et al., 1999;Ullerstam et al., 2003;Yang et al., 2016). These new bands remained when an argon blow-off process was carried out, suggesting that the surface-adsorbed sulfite or sulfate species between 1250 and 1000 cm −1 was chemisorbed (Zhang et al., 2006).
Upon adsorption of SO 2 on the surface of the NAu-2 sample before CP (Fig. 2c and d), the broad band from 3600 to 2800 cm −1 and the peaks at 1580 and 1675 cm −1 increased drastically with time. These absorbance bands were all attributed to the surface hydroxyl species (OH) and H 2 O. No peaks were observed over the range of 1000 to 1250 cm −1 , suggesting that the sulfite or sulfate products were not newly formed on the surface of the NAu-2 sample before CP. Upon adsorption of SO 2 on the surface of the NAu-2 sample after CP (Fig. 2d), the new bands at 3661 and 3450 cm −1 , the broad band between 3400 and 2700 cm −1 , and the broad band centered at 2131 cm −1 were observed as the exposure IMt-2 particles after CP were almost all less than 1 µm in size. Some Fe-rich particles with less Si and Al were observed on the nanoscale dimension. (c) NAu-2 particles with high Fe / Si ratios contain Mg, Al and Ca elements. (d) NAu-2 particles after CP were much smaller than the ones before CP. Some ferrihydrite clusters were observed and were attached on the surface of the NAu-2 particles after CP. (e) Typical SWy-2 particles were Al / Si rich, containing Fe, Mg and Ca elements. (f) TEM images of the SWy-2 particles after CP appeared smaller than the particles before CP. (g) The Si / Al-rich crystals in ATD particles were aluminosilicate with a low content of Fe, and typical α-Fe 2 O 3 particles were found to attach onto the aluminosilicate surface. (h) The pseudohexagonal nanoparticles were observed to on the surface of the α-Fe 2 O 3 crystal among the ATD particles. The SAED lattice constants of these nanoparticles were found to be very close to that of 2-line ferrihydrite. Figure 2. Comparison of the DRIFT spectra of mineral dust samples upon exposure to SO 2 for 45 min before and after CP. Data for IMt-2 (a, b), NAu-2 (c, d), SWy-2 (e, f) and ATD (g, h) are shown in the ranges of 4000 to 1250 cm −1 or 1250 to 1000 cm −1 . time increased. In detail, the band at 3661 cm −1 could be assigned to stretching vibration modes of isolated or bridged surface hydroxyl groups bonded to the surface iron ions embedded in the octahedral and tetrahedral sites (Faust et al., 1989;Nanayakkara et al., 2012;Zhao et al., 2015). The peaks at around 3450, 2131 cm −1 and the band between 3400 and 2700 cm −1 were all attributed to surface OH groups (Ma et al., 2010;Zhao et al., 2017). These new bands generated on the processed NAu-2 particles suggested that CP changed the location of diverse OH groups on the particle surfaces. Over the range of 1250-1000 cm −1 , the new bands centered at 1170 cm −1 were assigned to the asymmetric stretching of sulfate (Kong et al., 2014;Yang et al., 2015).
The spectra of the SWy-2 samples before and after CP ( Fig. 2e and f) showed a similar spectral character with those of the NAu-2 samples. The spectra for the ATD samples be- fore and after CP ( Fig. 2g and h) were roughly the same as the ones for IMt-2. All of the results demonstrated that the characteristic peaks for the active OH sites and the sulfite or sulfate products on the mineral particles after CP were significantly higher than those on the ones before CP, indicating the higher hygroscopicity and greater SO 2 uptake on the particles after CP. The data shown herein confirmed that CP could potentially promote the transformation of SO 2 on the particle surfaces.

Uptake coefficient of SO 2 on the mineral particles before and after CP
The areas of the bands (from 1250 to 1000 cm −1 ) attributed to the sulfite or sulfate products as a function of time are shown in Fig. 3. It was evident that the peak areas of the products on the mineral particles after CP were generally greater than the ones before CP. The reaction on the sample surfaces was practically saturated to SO 2 uptake within 15 min, except for the NAu-2 and IMt-2 samples after CP. As for all of the sample, the saturation coverages of the sulfite or sulfate products after CP were obviously greater than the corresponding values before CP, suggesting that CP favored the sulfate formation on the mineral surfaces due to improving active site number, as expected previously. The maximum uptake coefficients (γ geo and γ BET ) for SO 2 uptake on the samples were estimated on the basis of the sulfate formation rates in the initial 15 min. The values on the mineral samples before and after CP are shown in Table 1. The γ geo values of SO 2 on the IMt-2, NAu-2, SWy-2 and ATD samples before CP were 1.03 × 10 −7 , 0.30 × 10 −7 , 1.72×10 −7 and 1.37×10 −7 , respectively, which were in the order of SWy-2, ATD, IMt-2 and NAu-2. The γ geo values of SO 2 on the IMt-2, NAu-2, SWy-2 and ATD samples after CP were 4.7, 19.4, 2.7 and 2.0 times higher than the values before CP, respectively, suggesting that the SO 2 uptake on the mineral particles significantly increased after CP.
A BET was more appropriate to represent the effective area because the reactant may diffuse into tiny holes of the entire sample. The γ BET values of SO 2 on the IMt-2, NAu-2, SWy-2 and ATD samples before CP were 2.62×10 −12 , 0.75×10 −12 , 3.70 × 10 −12 and 1.61 × 10 −11 , respectively, which were in the order of ATD, SWy-2, IMt-2 and NAu-2. It was noteworthy that the S BET of samples increased after CP, as shown in Table 1. The γ BET values of SO 2 on the IMt-2, NAu-2, SWy-2 and ATD after CP were 2.2, 4.1, 1.5 and 1.4 times higher than the values before CP, respectively. The discrepancies in the γ BET value confirmed that the higher sulfate formation rates of the particles after CP were not only due to the increased surface area of the particles but also resulting from the chemical modification on the particle surfaces.
The estimated uptake coefficients were several orders of magnitude lower than the results from Ullerstam et al. (2003) and Usher et al. (2003), which could be partly explained by the difference in the preparation of mineral dust samples or the difference between diverse experimental structures such as the DRIFTS and Knudsen cell in kinetics discussion. In this study, mineral dust particles were in a highly accumulative state in the sample support of the Knudsen cell. The many layers of particles in the latter study will hinder the diffusion of gas into the underlayer particles, resulting in the underestimate of γ BET . However, the values herein were comparable to those obtained by the similar DRIFTS setup (Fu et al., 2007), indicating the reliability of our measurements.
After CP, the Fe A contents of the IMt-2, NAu-2, SWy-2 and ATD samples reached 1.8 %, 1.2 %, 1.7 % and 24.2 %, respectively, which increased by 2.6, 2.4, 2.4 and 6.4 times as compared to the ones before CP. The crystalline Fe (oxyhydr)oxides (Fe D ) contents of the samples after CP were not significantly changed as compared to the ones before CP, whereas the contents of the structural-Fe fraction in the Al-Si crystals of the IMt-2, NAu-2, SWy-2 and ATD samples after CP decreased by various degrees to 91.1 %, 96.1 %, 93.2 % and 42.5 %, respectively. Previous research had indicated that Fe A increased as a result of the simulated CP (Shi et al., 2009). Herein, we further proposed that the increased fractions of Fe A could be mostly transformed from the structural-Fe fraction in the aluminosilicate phase of the particles during CP, which is in good agreement with the TEM observation. For example, the Fe A in the ATD samples increased from 3.8 % to 24.2 % after CP, accompanied by a sharp decrease in the structural-Fe content from 60.7 % to 42.5 %.
The Mössbauer spectra and their fitted results are shown in Fig. 6. The corresponding hyperfine parameters estimated from the best fitted spectra are presented in Table S3. The central doublet with isomer shift (IS) of 0.37 mm s −1 and quadrupole shift (QS) of 0.72 mm s −1 were typical for highspin Fe(III) in octahedral symmetry (Eyre and Dickson, 1995), while the other one with IS of 1.12 mm s −1 and QS of 2.65 mm s −1 was characteristic of high-spin Fe(II) (Hofstetter et al., 2003;Kopcewicz et al., 2015). The two doublet components of the IMt-2, NAu-2, SWy-2 and ATD samples before CP were all attributed to different fractions of Fe(III) and Fe(II) in the aluminosilicate crystals. Before CP, the Fe(II) fraction in the IMt-2, NAu-2, SWy-2 and ATD samples were 34.0 %, 12.9 %, 18.3 % and 29.0 %, respectively (Fig. 6a, c, e and g). Furthermore, the spectra of the ATD sample before CP showed not only two central quadrupole doublets but also one MHS (magnetic hyperfine splitting) sextet with IS of 0.39 mm s −1 , QS of −0.13 mm s −1 and H f of 51.1 T. The MHS sextet, which shared 31.8 % of the whole area, could be ascribed to α-Fe 2 O 3 (Kopcewicz and Figure 6. Mössbauer spectroscopy measured for samples. IMt-2 before and after CP (a, b), NAu-2 before and after CP (c, d), SWy-2 before and after CP (e, f), ATD before and after CP (g, h). Experimental data were fit using a least-squares fitting program. The IS values were relative to α-Fe at RT. Prominent spectral features associated with different iron species are indicated. Kopcewicz, 1991), in agreement with the TEM analysis and free-Fe measurement as mentioned previously.
After CP, the Fe(II) contents of the samples decreased to 31.5 %, 11.6 %, 17.1 % and 10.9 %, respectively, for IMt-2, NAu-2, SWy-2 and ATD (Fig. 6b, d, f and h). It was supposed that the Fe(II) release is more energetically favorable than the one of Fe(III) due to the bond strength. As to the ATD sample after CP (Fig. 6h), not only did the Fe(II) fraction decrease from 29.0 % to 10.9 % but the Fe(III) frac-tion in the aluminosilicates also decreased from 39.0 % to 33.0 %. Meanwhile, the α-Fe 2 O 3 fraction was not significantly changed (31.8 % vs. 32.3 %). As discussed previously, the Fe mobilization was dependent on the specific chemical bonds. The Fe D phase in α-Fe 2 O 3 with the strong FeO bond was less liable than that embedded in the aluminosilicate lattice (Strehlau et al., 2017). It has been well documented that the Fe replacing alkaline elements as the interlayer ions was easily mobilized compared to the Fe bound by covalent bonds in the aluminosilicate matrix (Luo et al., 2005;Cwiertny et al., 2008;Journet et al., 2008). Therefore, the Fe in the aluminosilicate fraction of the mineral particles exhibited varied iron solubility. Particularly, a new quadrupole doublet with IS of 0.67 mm s −1 and QS of 1.21 mm s −1 was observed in the spectra of the ATD sample after CP (Fig. 6h), which shared 23.8 % of the total area, and was possibly indicative of the Fe(III) oxide hybridized in the aluminosilicate matrix (Kopcewicz and Kopcewicz, 1991). The free-Fe measurements have indicated that the Fe A fraction of ATD increased by 20.4 % after CP, so this Fe phase was most likely an amorphous Fe(III) hybridized with Al / Si. In terms of the other samples after CP, the magnetic signal of the newly formed Fe(III) phase was not detected. This was probably due to the fact that the newly formed Fe fractions were not available at sufficiently high levels to be clearly resolved by the Mössbauer spectroscopy method, and/or the slight signal drift and the poor signal-to-noise ratio made an unambiguous identification difficult. Herein, the newly formed amorphous Fe(III) phase was supposed to be a reactive Fe-bearing component, which may contribute significantly to the SO 2 uptake even at a low level.

The dissolution-precipitation cycle of the mineral Fe during CP
During the simulated CP experiments, the concentrations of total dissolved Fe (Fe s ), dissolved Fe(II) and Fe(III) released from the particles, as a function of time are shown in Fig. 7. Similar dissolution trends were observed for all of the samples. One can see that the suspensions at pH 2 induced a rapid increase in Fe s . Increasing the pH from 2 to 5 resulted in a rapid and almost complete removal of Fe s . In fact, only a rather small fraction of the Fe in dusts could be dissolved at pH values above 4 (Zuo and Hoigne, 1992). The dissolved Fe precipitated rapidly as an insoluble deposit at pH 5. When the suspension pH was again reduced to 2, a steep increase in the Fe s concentration was measured once again. The fast Fe release was due to the redissolution of the Fe-rich precipitates, which was proposed to be reactive Fe phases (Shi et al., 2009(Shi et al., , 2015. Such highly soluble Fe-bearing precipitates Figure 8. TEM images of the newly formed particles in the precipitation experiment. Based on the TEM-EDX measurement and SAED analysis, these particles could be categorized into two different types, which are circled in (a). The typical sizes of the first type were hundreds of nanometers. The enlarged images are displayed in (b, c, d). The insert EDX data and SAED image confirmed that they were poorly crystalline aluminosilicate with low Fe but high Si / Al contents. The second particle type (e, f, g) was Fe rich but with a lower amount of Si / Al, which was nearly 1 µm in size. Based on the EDX data and the SAED analysis, these bigger particles were ambiguously identified as Na 0.42 Fe 3 Al 6 B 309 Si 6 O 18 (OH) 3.65 . Figure 9. Mössbauer spectroscopy measured at RT for the newly formed particles collected in the precipitation experiment.
have been observed under the TEM, as well as the free-Fe measurement and Mössbauer characterization. For each pH cycle during the simulated CP experiment, the overall changes in total released Fe concentrations were reproducible. The Fe ion on the particle surfaces would experience a continuous process of dissolution, precipitation, redissolution and reprecipitation when the pH cycles between pH 2 and pH 5 (cloud-aerosol modes). During this process, the Fe(II) fraction would be transformed to Fe(III). The results shown herein suggest that CP could significantly modify Fe partitioning between dissolved and particulate phases in the real atmosphere. Not only did the increase in specific surface area contribute to the enhanced sulfate formation but the highly reactive Fe on the particle surfaces yielded during CP were also responsible for the higher SO 2 uptake on the particles after CP.
When investigating the NAu-2 sample, once the pH of the clear solution increased from 2 to 5-6, the Fe-bearing nanoparticles separated out from the solution rapidly and precipitate out slowly. The sample developed an initial yellow color and then an orange colored suspension. The TEM images of the precipitated particles are shown in Fig. 8. The particles could be categorized into two different types. One type of particle could be characterized as hundreds of nanometers in size, with low Fe but high Si / Al content. The other type displayed particle sizes of nearly 1 µm and were Fe rich but contained a smaller amount of Si / Al components. These bigger particles were ambiguously identified as Na 0.42 Fe 3 Al 6 B 309 Si 6 O 18 (OH) 3.65 on the basis of the EDX data and SAED analysis. It is likely that the Al / Si elements also precipitated out along with the Fe.
The Mössbauer spectra of the precipitated Fe-rich particles are shown in Fig. 9 (Kopcewicz et al., 2015). The results were in good agreement with the TEM observation, which showed that most of these Fe particles were mostly present as the Fe(III) hybridized with Al / Si. The particle size distributions in the suspensions were also determined by dynamic light scattering, as shown in Fig. 10. When pH was lower than 2.0, the particles seemed to stabilize below 10 nm in size. These Fe colloids were thought to be a source of soluble Fe (Janney et al., 2000). Once pH increased, the size of precipitated particles quickly increased, even to the microscale, and the suspension was featured with a polydispersed size distribution. Conclusively, the precipitated Fe particles were mainly Fe(III) with weak crystal structure and/or ferrihydrite nanoparticles hybridized with Al / Si, which were possibly transformed from the Fe hosted in the aluminosilicate matrix of the particles. The particle surfaces after CP were coated by these reactive Fe to provide more surface OH species, resulting in enhanced SO 2 uptake.

Conclusion and implication
Transition metal ions, especially Fe(III), can catalyze SO 2 oxidation rapidly in cloud drops (Harris et al., 2013). This study further confirmed that SO 2 uptake on the mineral particles could be greatly enhanced by CP, possibly more than described previously. The higher uptake coefficient of the particles after CP was not only due to increased surface area but also resulted from the chemical modification of the particle surfaces. The free-Fe and Mössbauer analyses suggested that CP triggered new formation of amorphous Fe particles on the surfaces, which were mostly transformed from the Fe hosted in the aluminosilicate matrix. TEM showed that the amorphous Fe(III) and/or ferrihydrite nanoparticles were hybridized with Al / Si. In general, the acidity fluctuation during CP enables the dissolution-precipitation cycles of mineral Fe to yielded more reactive Fe, resulting in more SO 2 uptake on the particle surfaces. More SO 2 adsorption further increases the surface acidity of dust particles, in turn leading to higher Fe solubility; again, more sulfate formation. It was thus proposed that there is a positive feedback relative to SO 2 uptake and iron mobilized from mineral particles during CP, therefore enhanced sulfate formation greatly.
Our results also serve to explain high sulfate loading in fog-haze episodes of China. It has been recommended that sulfate contributed significantly to the explosive growth of fine particles, thus exacerbating severe fog-haze development (Kasibhatla et al., 1997;Nie et al., 2014;Barrie et al., 2016). Haze and fog within an episode was often found to transform each other at a short time due to the diurnal variation in RH, whereby the haze-fog transition was probably analogous to the aerosol-cloud interaction. Water content of aerosol or fog drops was regulated by RH, and thus allowed the particle acidity fluctuation. Although the aerosol acidity could not be accurately determined from field measurements or calculated using the thermodynamic model, we recognized that the large pH fluctuations between the haze-fog modes could significantly modify the microphysical properties of mineral particles, trigger formation of reactive Fe particles and thus accelerate sulfate formation via a selfamplifying process, contributing to explosive growth of fine particles at the initial stage of fog-haze events. The data presented herein also highlight that CP provides more bioavailable iron from mineral particles than one expected previously, which is a key speciation to promote oceanic primary productivity. Results of this study could partly explain the missing source of sulfate and improve agreement between models and field observations. Additionally, previous studies indicated that Fe in pyrogenic aerosols was always presented as liable Fe, such as ferric sulfate and aggregated nanocrystals of magnetite (Fe 3 O 4 ) (Fu et al., 2012), and displayed higher Fe solubility compared to dust (Desboeufs et al., 2005;Sedwick et al., 2007;Ito et al., 2019). Alexander et al. (2009) demonstrated that the sulfate formed through metal catalysis was highest over the polluted industrial regions of northern Eurasia, suggesting that the increasing importance of the metal-catalyzed S(IV) oxidation pathway due to anthropogenic emissions (Alexander et al., 2009). With the rapid development of industry and agriculture, the pyrogenic Fe-containing aerosols are indispensable contributors to the atmospheric Fe load in China. Thus, the acidic solution at pH 2 and high sulfate loading of fine particles in severe fog-haze events of China might be more relevant to Fe-containing combustion aerosols than mineral dust. Based on the current findings, not only the po-tential influences of cloud liquid water content, light and organic ligands but also the solubility and speciation of Fe in pyrogenic aerosols will be considered during the simulated CP experiments in the future. A more detailed understanding of the iron-sulfur cycle during CP is therefore critical to estimate accurately the contribution of CP to global sulfate loading and its impact on the climate. Data availability. All data described in this study are available upon request from the corresponding authors.
Author contributions. ZW, HF and JC designed the experiments. ZW, TW, HF and LZ performed the laboratory experiments. HF, JC, LZ and VG contributed reagents and analytic tools. CG, VG and MT gave some valuable suggestions in designing the experiments. ZW, TW and HF analyzed data. ZW and HF wrote the article with inputs from all coauthors.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Multiphase chemistry of secondary aerosol formation under severe haze". It is not associated with a conference.