A Novel Pathway of Atmospheric Sulfate Formation Through Carbonate Radical

Shanghai Key Laboratory of Atmospheric Particle Pollution and Prevention, Department of Environmental Science and Engineering, Fudan University, Shanghai, 200433, P. R. China. Shanghai Institute of Pollution Control and Ecological Security, Shanghai, 200092, Peoples’ Republic of China. Key Laboratory of Cluster Science, Ministry of Education of China, School of Chemistry and Chemical Engineering, Beijing Institute of Technology, Beijing 100081, P. R. China Univ. Lyon, Université Claude Bernard Lyon 1, CNRS, IRCELYON, F-69626, Villeurbanne, France.


Supplement Text 4
Determination of reaction order. Calculation of uptake coefficient is the way to describe the reaction efficiency of gas/surface interactions. Before that, sulfate formation rates as a function of SO2 concentration were initially determined to verify its reaction order in the selected concentration range (Fig. S9 A-E), which is a crucial step to give a credible estimation of SO2 uptake coefficient. Based on a prior study (Shang et al., 2010), SO2 uptake on particle surfaces depends on the SO2 concentrations and active sites, which thus could be described by the following equation (Eq. S1):  ( [CO2]) to SO2 (CO2) gas concentration employed in the system, [TiO2] to the concentration of active sites on the TiO2 particle surfaces, and [H2O] represents for surface water concentration, and m, l, n, and p are the reaction orders of corresponding species. Clearly, the steady growth of sulfate on TiO2 particles in the presence of 9.83×10 15 molecules cm -3 CO2 made a clear indication (panel A-E) that the decrease in surface active sites is negligible. Meanwhile, mass flow controllers provide stable gas flow and maintain the constant concentrations of humidified air and CO2, which allow us to simplify the Eq. S1 to Eq. S2 through a logarithm function. for the concentration of SO2 where particles are exposed. We then plotted the sulfate formation rate against exposed SO2 concentration. Linear fitting analysis for those points resulted in 1.13 order for the reaction, with R 2 ≈ 0.99 (Fig. S9 F). So far, 400 ppb is the lowest concentration that we are able to apply for the uptake measurements due to the limitation of the current experimental setup. The prior work has demonstrated that atmospheric SO2 concentration reaches up to 40 ppb (Franchin et al., 2015). We note that a difference of nearly a factor of 10 remains for the SO2 concentration employed in laboratory studies and that measured in field observations. Nevertheless, we have already verified its pseudo-first-order kinetic in the wide range of 400-20000 ppb. This leads us to assume that uptake coefficients estimated under ppm level remain valid, and those datasets derived from laboratory chambers are able to be generalized to the atmosphere condition.

Supplement Text 5
Methodology for uptake coefficient estimation. The reaction uptake coefficient was estimated by the following Eqs. S3-5, as suggested by the previous work (Kong et al., 2014): where ν is the mean molecular velocity of SO2, As the effective sample surface, R the gas constant, T the temperature, MSO2 the molecular weight of SO2, and a total number of surface collisions per unit time (Z). To be precise, the formation rates (d[SO4 2-]/dt) in the equation were determined by IC measurements, followed by a conversion factor calculation through linear regression analysis for the integrated absorbance of sulfate bands and corresponding sulfate concentrations. By employing this method, a conversion factor of f = 6.34×10 15 ions integrated absorption units −1 was obtained and corresponding SO2 uptake coefficients in the TiO2+ (CO2) + SO2 system were thus estimated. For SO2 uptake in the TiO2-(CaCO3) + SO2 system, we estimated SO2 uptake coefficients using three conversion curves established for various types of dust particles. For this sake, we mixed known proportion of K2SO4 and selected dust particles, and thus obtained relationship curves between the integrated absorbance of DRIFTS sulfate bands (1300-1100 cm -1 ) and corresponding theoretical sulfate ions through linear fitting (Fig. S4).

Supplement Text 6
Gas sources and configuration setup. All gases used for experiments were purchased from Shanghai TOMOE gases CO., LTD. SO2 (a mixture of 2.46×10 15 molecule cm -3 SO2 and balance gas N2), CO2 (a mixture of 2.46×10 17 molecule cm -3 and balance gas N2 or pure CO2), and air with ultra-high purity of 99.999% were used as received. To precisely control the gas flow and concentrations, mass flow controllers (Beijing Sevenstar Electronics Co., Ltd.) were employed to adjust SO2 or CO2+SO2 flows to the required concentration. With regard to the relative humidity (RH) adjustment, one pure dry air (21% O2 and 79 % N2) flow was humidified in a bubbler loaded with ultrapure water (electrical resistivity=18.2 MΩ). Together with rest gas flow equipped with MCFs, mixed gas with desired RH could be achieved. However, it should be noted that gas flow should be mixed for at least 1 h to reach a stable status before it is allowed to be introduced into DRIFTS or a home-made quartz chamber (d= 5.5 cm, H=1.2 cm, and actual volume= 20 cm 3 ).

Supplement Text 7
Mineral dust pretreatment and characterizations. Degussa TiO2 was used as reference dust particles in all tests. A pretreatment process was performed to remove the organic impurities and inorganic impurities such as bicarbonate and carbonate at TiO2 particle surfaces using a thermal heating method with a moderate heating rate of 2 °C min -1 at 400 °C for five hours in a tube furnace with the protection gas flow (dry air, 100 mL min -1 ). Similar to the procedures applied for the particles in TiO2+CO2 +SO2 system, particles TiO2 and CaCO3 (Aladdin industrial Co., LTD) in the TiO2-(CaCO3) + SO2 system were separately calcined at 500 °C for 3 h with protection gas flow (dry air, 100 mL min -1 ). Similarly, SiO2 and Al2O3 (Aladdin industrial Co., LTD) went through heat pretreatment described above (500 °C for 3 h). Mixtures of SiO2, Al2O3, CaCO3 and TiO2 were prepared by manually grinding in the agate mortar. The crystalline phase of TiO2 dust powder was analyzed by diffractometer with a Cu-Ka source. In addition, Brunauer-Emmet-Teller (BET) adsorption isotherm measurements were as well conducted using a Microm-eritics Asap 2000 surface area analyzer.

Supplement Text 8
In situ DRIFTS measurements. To investigate the surface species during the heterogeneous reactions of SO2 on mineral dust particles TiO2, in situ diffuse reflectance Fourier transformed infrared spectroscopy (in situ DRIFTS, IRTracer-100, Shimadzu Instrument Corporation) was applied. Those FTIR spectra ranging from 4500 to 700 cm −1 were collected by using high-sensitivity mercury cadmium telluride (MCT) detector with a resolution of 4 cm −1 for 100 scans and recorded by diffuse reflectance accessory. Thermocouple wires attached to the ceramic sample holder and a temperature controller were used, allowing resistive heating from simulated solar light (250 W xenon lamp, Beijing CEAULIGHT Co.Ltd., China) during the reaction. Moreover, a cooling circulation device was applied considering the temperature fluctuation of samples due to the long-term exposure to the light source. All measurements were carried out at 298 K. Prior to reaction gas entering into the cell, further pretreatment was performed by sending dry airflow to the DRIFTS chamber under light irradiation for 1 h to remove the contamination residual on particle samples. S4 Supplement Text 9 IC measurements. Sulfate ions formed on dust particles during the reaction were investigated by IC using a Metrohm 883 Basic IC system equipped with a Metrosep A supply 5-250/4.0 analytical column and a conductivity detector. Sample particles after reaction were extracted by an oscillator. The leaching solution (3 % formaldehyde solution applied for TiO2+ (CO2) + SO2 and 2 % isopropanol that applied for TiO2-CaCO3 +SO2 were prepared using ultrapure water of 18.2 Ω). These solutions served as a preservative to avoid SO2 (ad) oxidation during the extraction through a 0.22 µm PTFE membrane filter, followed by IC measurement. In this work, we employed a scrubber system with an eluent of 3.2 mM Na2CO3/1.0 mM NaHCO3 for anion measurement at a flow rate of 0.7 mL min −1 .

Supplement Text 10
Raman spectra. In this study, an XploRA confocal spectrometer (Jobin Yvon, Horiba Gr, France) was applied for investigating dust particles after reaction. The Raman scattering was excited by an external-cavity diode (532 nm) and coupled with a 50× Olympus microscope objective (Olympus, 0.50 Numerical Aperture). Raman signals were acquired by a charge-coupled device (CCD) camera operating at 220 K with 1200 lines per mm diffraction grating. Samples after reaction were monitored at the range of 100-2000 cm -1 with twice repeated scans and 5 s exposure time per spectrum to give an overall view of mineral dust particles, carbonate, and formed sulfate species. A narrowing scanning from wavenumber 910 to 1050 cm -1 , with 40 s exposure time and 15 times repeated scans, was conducted to investigate sulfate species produced on dust particle surfaces.

Supplement Text 11 ESR measurements.
Electron spin resonance (ESR) spectrometry was performed on a JES-FA200 Spectrometer and formed radicals using DMPO spin trap were recorded at room temperature for all measurements. The g factor in each data was calibrated using manganese standards. ESR spectrums were recorded after irradiation with a 500 W high-pressure mercury lamp (USH-500SC) for 2 min. The computer simulations of ESR spectrums were further conducted with the software Isotropic Radicals (Hagen, 2009) to identify active intermediates during the reaction course. For details, computer simulation of the four-line DMPO/· OH and six-line DMPO/SO3 .radical adduct were obtained by setting splitting constant of a N =a β H ; = 15.2 G, and a N =14.7 G and a β H =16.05 G, respectively. Those employed constants compare favorably with the prior literature (Mottley and Mason, 1988).

Supplement Text 12 NTAS Measurements.
We utilized a nanosecond transient absorption spectroscopy (NTAS), designed and made by Dalian Institute of Chemical Physics, to investigate the possible heterogeneous reaction channels of sulfate formation on dust particles in the presence and absence of carbonate ions. About 35 mJ of the Nd: YAG laser source (355 nm, laser pulse 9 ns) entered into suprasil self-masking cuvettes (1 cm optical path and actual volume= 4 mL). All transient absorption spectra averaged by 10 times parallel results were obtained at room temperature. In addition, a monochromator and a photomultiplier tube with a detection wavelength range from 200 to 800 nm enable us to monitor the major active intermediates in the mineral dust suspension system. To visualize the signal of carbonate radical, we here applied single exponential decay curves (Eq. S6, the left side of the forward slash) to fit the signal of CO 3 •evolved from TiO2 suspension containing carbonate ions. The consumption process of carbonate radicals in chain reactions is described by a single exponential growth curve (Eq. S6, the right side of the forward slash).
where is the lifetime of carbonate radicals in each system. Constant functions were employed to fit the samples without NTA signals.

Supplement Text 13
Theoretical studies of two reaction routes for producing carbonate radicals. In the present system, one of the viable mechanisms for the formation of CO 3 •is hydrogen atom transfer (HAT) reaction (Eq. S7) through •OH with HCO3 -. Such optimized structures involved in the HAT process as a reactant (React), transition state for the reaction step (TS), and the resulting product (Prod) were displayed in Fig. S11A. In HAT reaction, HCO3transfers a H atom to O atom in •OH, leading to the formation of H2O and CO 3 •-. The corresponding calculated activation free energy barrierΔG ≠ HAT was 9.02 kcal mol -1 , which is low enough for the HAT reaction of •OH and HCO3to occur rapidly. The rate constant kHAT-1 predicted for the HAT reaction was 4.04×10 6 M -1 s -1 .
In addition, the formation of CO 3 •could be also through the SET reaction (Eq. S8) of •OH with CO3 2-. As shown in Fig.  S11B, O atom in CO3 2transfers an electron to O atom in •OH to form CO 3 •and OH -. This SET reaction is thermodynamically favorable, with the difference of Gibbs free energy between reactant and product ΔG 0 SET lying at -9.74 kcal mol -1 . The corresponding calculated activation free energy barrier ΔG 0 SET is 0.02 kcal mol -1 , which is low enough for S5 this thermal process to be faster than that controlled by the diffusion. Thus, the SET process for •OH and CO3 2is diffusion-controlled. The rate constant kSET-2 predicted for SET process between •OH and CO3 2was 7.65×10 9 M -1 s -1 .
Supplement Text 14 Methodology for DFT calculations. All electronic calculations were performed using Gaussian 09 program (M. J. Frisch, 2009). In recent years, numerous benchmarks concerning the performance of different Density Functional Theory (DFT) methods have proved that M06-2X functional is a reasonable choice for computing thermochemistry and barrier of atmospheric relevant systems (Zhao and Truhlar, 2008;Elm et al., 2012;Mardirossian and Head-Gordon, 2016;Pereira et al., 2017). Hence, geometry optimization of all the investigated species, including reactants, products, and transition states and frequency calculation have been performed using the M062X functional coupled with the 6-311++G(3df, 3pd) basis set. Single point energy was calculated at CCSD(T)-F12/cc-PVDZ-F12 level of theory by ORCA 4.10 program using the geometry optimized at M06-2X/6-311++G(3df, 3pd) level (Neese, 2012;Peterson et al., 2008;Yousaf and Peterson, 2008;Lane and Kjaergaard, 2009;Knizia et al., 2009). In order to mimic the implicit solvent environment (Marenich et al., 2009), the solvation effect was introduced through the continuum solvation model based on solute electron density (SMD). The free energy in the solution was obtained using Eq. S9: E 0 CCSD(T)-F12 is the electronic energy obtained at the CCSD(T)-F12/cc-PVDZ-F12 level, G corr M06-2X is the thermal correction to Gibbs free energy at the M06-2X/6-311++G(3df, 3pd) level under standard conditions (T=298K, P=1 atm), G SMD M06-2X is the solvation free energy obtained from single point calculation at the M06-2X/6-311++G (3df, 3pd) level and G 1atm→1M is an additional term for converting the standard pressure of 1 atm in the gas phase to the standard concentration of 1 mol L -1 in solution (1.89 kcal mol -1 ) (Marenich et al., 2009).
In the present system, one of the viable mechanisms to describe reaction process is the hydrogen atom transfer (HAT) reaction. In this kind of reaction, transition states are necessary for calculating the activation free energy barrier term in HAT reaction. kact, the thermal rate constant determined by activation free energy barrier, was calculated using the conventional Transition State Theory (TST) (Eyring, 1935;Evans and Polanyi, 1935;Truhlar et al., 1983) as implemented in 1 M standard state (Eq. S10): Where l is the reaction path degeneracy accounting for the number of equivalent reaction paths. κ(T) , a temperature-dependent factor, corresponds to quantum mechanical tunneling and is approximated by Eq. S11 according to Wigner (Elm et al., 2013;Wigner, 1932) for the HAT reaction; kB is Boltzmann's constant, h is Plank's constant, and R is gas constant.
However, for a single electron transfer (SET) reaction, transition state cannot be located using electronic structure methods, as it is not possible to describe the mechanistic pathway of electron motion. To estimate the reaction activation free energy barrier (ΔG ≠ term), in such cases, the Marcus theory was used (Marcus, 1997a, b). Within this transition-state formalism, the SET reaction activation free energy barrier (ΔG ≠ SET ) is defined in terms of the difference of the Gibbs free energy between reactants and the products (ΔG 0 SET ) in SET reaction and the nuclear reorganization energy (λ) as follows (Eq. S12)： [S12] While for λ, a very simple approximation was used as Eq. S13: Where E is the non-adiabatic energy difference between reactants and vertical products. It is expected that the tunneling effect would not be large for the reactions with low barriers (Iuga et al., 2015;Luo et al., 2017). Therefore, the tunneling effect was neglected in this SET reaction for its low barriers which will be discussed in detail in the R&D Section. In addition, diffusion should be considered especially for the low barrier SET reaction with kact approximated to the diffusion limit (Luo et al., 2017). Thus, the Collins-Kimball theory (Kimball, 1949) was used to calculate the total rate constant kSET or kHAT that can be expressed as Eq. S14: Where kD is the steady-state Smoluchowski rate constant (Smoluchowski, 1917) for an irreversible bi-molecular diffusion-controlled reaction. Smoluchowski (Raymond, 2000) developed an empirical equation to calculate diffusion-controlled reaction rate constant kD in an aqueous solution (Eq. S15): Where NA is the Avogadro constant, kB is Boltzmann constant, T is temperature, η is the viscosity of water, and r1, r2 are the radius of the bi-molecular reactants.

Supplement Text 15
The procedures for oxygen isotope experiments: Oxygen isotope experiments were divided into two parts: 1. preparation of C 18 O2/N2 (2.46×10 17 molecule cm -3 ) through C 16 O2 exchanging with H2 18 O, 2. monitoring formed species using in situ DRIFTS. In the first step, a chamber was washed with ultrapure water (electrical resistivity= 18.2 MΩ), followed by drying the chamber using dry ultrapure N2 gas flow (300 mL min -1 , 40 min). Afterward, we introduced H2 16(18) O into the chamber with a pipette and operated this process in the glove box with N2 atmosphere, where contents of H2O were rigorously controlled below 2.46×10 13 molecule cm -3 to prevent the undesired exchange of 18 O with 16 O in the atmosphere. Followed by this step, we extracted the gas from the chamber and added it with CO2 for O exchange (70 min). For these two sets of DRIFTS studies, TiO2 particles went through in situ heat pretreatments in the sample holder with an airflow (150 mL min -1 ) at 500℃ for 0.5 h. Subsequently, TiO2 particles were cooled down to room temperature using a cooling device attached to the DRIFTS for 35 min. Then, the chamber was sealed with gas-tight sample pads and we evacuated the chamber and introduced C 16(18) O2 /N2. After adsorption of CO2/N2 for 20 min, TiO2 was exposed to 6.14×10 15 molecule cm -3 of SO2/N2+O2 (RH= 30 %) for 60 min and we monitored the sulfate formation under irradiation (light intensity (I) = 30 mW cm -2 ) for each set of experiment.

Supplement Text 16
Sampling. Our sampling for atmospheric particle matter was launched on the roof of the environmental science and engineering department, Fudan University (Jiangwan Campus, 31.340661°N, 121.506747°E, 16 km away from the city downtown. Observations for water-soluble ionic components of particle matter were performed using an 8-stage non-viable-cascade-impactor type sampler (TISCH TE Inc., USA), size gradient of which are in the sequence of < 0.43, 0.43-0.65, 0.65-1.1, 1.1-2.1, 2.1-3.3, 3.3-4.7, 4.7-5.8 5.8-9.0 and 9.0~ μm. These sizes represent the effective cutoff diameter at each level for unit density spherical particles. In our sampling, the estimations for (bi)carbonate ions in atmospheric particulate matter mainly focus on the initial four stages (3.3~ μm). Atmospheric airflow from the head was maintained at the constant 28.3 L min -1 to meet the operation required for the Anderson-type sampler configuration. Quartz filters (81 mm in diameter, Whatman, GE Healthcare, UK) were employed for sampling operations, and membranes were pretreated with ultrapure water (electrical resistivity= 18.2 MΩ) no less than three times (kept in the ultrasonic cleaning tank for 40 min and then rinsed with ultrapure water for twice) before being sent into the infrared drying oven, followed by packing them in aluminum foils prior to field sampling. We carried out this procedure to eliminate the water-soluble background ions as much as possible and to ensure the balance calculation for (bi)carbonate ions. Additionally, blank membranes following the same aforementioned pretreatments were applied for measuring the background concentrations for ions. We separated our daily samplings into two periods, 11 hours for each, to give an insight into the influence of potential photo-induced reactions on secondary sulfate formation in the atmosphere.

Supplement Text 17
Chemical analyses. Water-soluble ions in the atmospheric particulate matter were measured using a Metrohm 883 Basic IC system, which is equipped with a Metrosep A Supply A-250/4.0 anion analytical column or a C-150/4.0 cation analytical column and a conductivity detector; eluent: 3.2 mM Na2CO3/1.0 mM NaHCO3 for anion measurements at a flow rate of 0.7 mL min −1 ; 1.7 mmol L -1 HNO3 for cation measurements at a flow rate of 0.9 mL min −1 ; inject volume: 20 uL. A series of Li + , NH4 + , Ca 2+ , Mg 2+ , Na + , K + , F -, CH3COO -, COOH -, Cl -, NO2 -, NO3 -, PO4 3-, SO4 2-, and C2O4 2ions were analyzed. Each collected filter membrane was cut and placed into a 50 mL PET vial, followed by ultrasonically extracting for 40 min with 20 mL ultrapure water (R ≥ 18.0 MΩ). For each sample, we applied two PTFE filter chips to remove insoluble particulate matter before being sent into IC. Obtained ion concentration results were further transformed to the data in the unit of μg m -3 on the basis of Eq. S16: where ρ represents the ions concentration (μg m -3 ), ρ 1 is the measured ions concentration of membrane after sampling (mg L -1 ), ρ 0 is the measured ions concentration of membrane before sampling (mg L -1 ), V is the volume of extracting solution for sample (L), and N refers to divisions for membrane, D to the dilution coefficient, and Vnd to the total sampling volume.
Supplement Text 18 (Bi)carbonate estimation. Two assumptions were made to simplify the estimation for bi(carbonate) ion concentrations within the system: a. dominating cation ions in each system are H + , Li + , Na + , NH4 + , K + , Mg 2+ and Ca 2+ whereas those transition and heavy metal ions were out of consideration given their limited contents of dissolve cations in the atmospheric particulate matter; b. three typical organic acid ions (CH3COO -, COOHand C2O4 2-, major soluble organic acid ions in the atmosphere) were taken into accounts for ionization balance and the rest of the charge gap in each system were assumed to originate from (bi)carbonate ions. Then we established balance equations (Eqs. S17-20) for each sample on the basis of charge conservation and ionization equilibrium constant of carbonic acid (K1= 4.47×10 -7 and K2= 4.69×10 -11 at 273 K).
[ It worth mentioning that [X] is referred to charge concentration for ions (Coulomb•M). Additionally, the temperature factor was also considered to correct equilibrium constant for (bi)carbonate ions using the following equation: where ∆H is the temperature variation (K), R the ideal gas constant (8.31451 J· mol -1 ·K -1 ), and T the temperature (K) during pH measurements. We then solved those equations to obtain a series of [HCO 3 -], which were finally corrected by subtracting blank values.