A novel pathway of atmospheric sulfate formation through carbonate radicals

Carbon dioxide is considered an inert gas that rarely participates in atmospheric chemical reactions. Nonetheless, we show here that CO₂ is involved in some important photo-oxidation reactions in the atmosphere through the formation of carbonate radicals (CO${}_{\mathrm{3}}^{\mathrm{•}-}$). This potentially active intermediate CO${}_{\mathrm{3}}^{\mathrm{•}-}$ is routinely overlooked in atmospheric chemistry concerning its effect on sulfate formation. The present work demonstrates that the SO₂ uptake coefficient is enhanced by 17 times on mineral dust particles driven by CO${}_{\mathrm{3}}^{\mathrm{•}-}$. Importantly, upon irradiation, mineral dust particles are speculated to produce gas-phase carbonate radical ions when the atmospherically relevant concentration of CO₂ presents, thereby potentially promoting external sulfate aerosol formation and oxidative potential in the atmosphere. Employing a suite of laboratory investigations of sulfate formation in the presence of carbonate radicals on the model and authentic dust particles, ground-based field measurements of sulfate and (bi)carbonate ions within ambient PM, together with density functional theory (DFT) calculations for single electron transfer processes in terms of CO${}_{\mathrm{3}}^{\mathrm{•}-}$-initiated S(IV) oxidation, a novel role of carbonate radical in atmospheric chemistry is elucidated.

1 Introduction

Atmospheric composition changes are subjected to highly reactive light-induced radicals, such as hydroxyl (•OH), hydroperoxyl (HO₂•), or nitrate radicals (NO₃•), which are able to alter not only compositions but also the physical and chemical properties of particulate matter (T. Liu et al., 2022; Stevenson et al., 2020; Mahajan et al., 2021). However, when atmospheric chemical reactions occur over humified particles at ambient conditions, which creates a locally enriched aqueous medium of unique chemical activity, other radicals might likewise gain importance. The carbonate radical (CO${}_{\mathrm{3}}^{\mathrm{•}-}$) is typically such an active radical. The lifetime of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ ranges from a microsecond to even a few milliseconds, and its concentration can be 2 orders of magnitude higher than that of hydroxyl radicals over the water surface (Chandrasekaran and Thomas, 1983; Goldstein et al., 2001; Shafirovich et al., 2001; Sulzberger et al., 1997). In addition, the one electron reduction potential of the E⁰(CO${}_{\mathrm{3}}^{\mathrm{•}-}$ $/$ CO${}_{\mathrm{3}}^{\mathrm{2}-}$) coupling is 1.78 V vs. a normal hydrogen electrode (NHE) at neutral pH, leaving CO${}_{\mathrm{3}}^{\mathrm{•}-}$ a strong oxidant in aquatic chemistry (Cope et al., 1973; Bisby et al., 1998; Merouani et al., 2010). Previous studies concerning carbonate radical in aqueous media demonstrate that it reacts rapidly with some organic compounds (Merouani et al., 2010), especially for those electron-rich amine compounds (Stenman et al., 2003; Yan et al., 2019). Also, it has been pointed out that the scavenging of hydroxyl radicals by (bi)carbonate species leads to the formation of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ ions (Graedel and Weschler, 1981; Xiong et al., 2016). Besides, a high second-order rate constant, lying at 10⁹ M⁻¹ s⁻¹, has been reported for the reaction of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ with porphyrins (Ferrer-Sueta et al., 2003), indicating that this radical ion has great oxidation capability that may trigger atmospherically relevant chemical reactions, i.e., secondary inorganic species formation. However, it is only regarded as a marginal intermediate in tropospheric anion chemistry so far (Lehtipalo et al., 2016; Beig and Brasseur, 2000; Herrmann et al., 2000; Ge et al., 2021), and its underlying role as an active oxidant for heterogeneous reaction in the atmosphere is barely explored. Very recently, our group observed the promotional effect of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ on atmospheric nitrate formation (Fang et al., 2021). Motivated by this finding, attempts were made to further explore its role in other important, atmospherically relevant reactions.

It is well documented that sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$) is also a key constituent of aerosols in the atmosphere (Huang et al., 2015; Su et al., 2016). It is able to serve as the precursor of efficient cloud condensation nuclei, with optical properties leading to a cooling effect (Wang et al., 2011). As a consequence, the mechanism aspect of secondary sulfate formation was the focus of numerous studies over the past decades (Hung et al., 2018; Stone, 2002; Zheng et al., 2015). There is a consensus that high-valence sulfur (VI), produced from the oxidation of anthropogenic SO₂, is the dominant source of atmospheric secondary sulfate. However, a remarkable missing sulfate budget emerges for the atmospheric modeling (Huang et al., 2019; Liu et al., 2021; Itahashi et al., 2018), which significantly underpredicts SO${}_{\mathrm{4}}^{\mathrm{2}-}$ with respect to observational results when heterogeneous aerosol chemistry is not considered (Zheng et al., 2015; Feng et al., 2018; Wu et al., 2021). This indicates that the heterogeneous sulfate production pathway is a crucial process, and exploring the unrecognized heterogeneous mechanism is very likely to narrow the gap between observations in lab studies, field measurements, and numerical modeling. Due to the missing chemical mechanism that initiated a fast SO₂ oxidation, atmospheric models fail to capture the key feature of atmospheric observations of high sulfate production during dust storm episodes in the troposphere (Yu et al., 2020; Dong et al., 2016; Huang et al., 2014), where an evident increase in Ca²⁺ (Wang et al., 2005; Li et al., 2013), carbonate-containing particles with high alkalinity (Tang et al., 2016; Li et al., 2014; Abou-Ghanem et al., 2020), and photoactive mineral components (Nie et al., 2012; Ta et al., 2003) are prevalent. Air mass is usually in low relative humidity, reportedly being 25 %–35 % (Al-Salihi and Mohammed, 2015; Csavina et al., 2014; Najafpour et al., 2020) in these events, during which the photochemical process is able to alter atmospheric constituents (X. C. Liu et al., 2022). Consequently, there are unknown heterogeneous reaction pathways and previously unconsidered promoters that have great potential to accelerate sulfate formation in the dust-storm-relevant conditions.

Due to the high stability of CO₂ under ambient conditions (Hossain et al., 2020), there are rare studies concerning the influence of CO₂ in atmospheric chemical processes (Deng et al., 2020; Y. Liu et al., 2020; Xia et al., 2021). CO₂ is demonstrated to form (bi)carbonate species over humidified dust particles (Baltrusaitis et al., 2011; Nanayakkara et al., 2014) and is reduced to CO under solar illumination (Deng et al., 2020). Nonetheless, its impact on atmospheric heterogeneous reactions remains poorly characterized. Our early laboratory study illustrates that CO₂ decreases the sulfate formation on aluminum oxide particles in the dark (Y. Liu et al., 2020), while upon solar illumination its role in SO₂ oxidation over mineral dust surfaces is still an open question. In addition, carbonate salt is abundant in authentic dust particles (Cao et al., 2005) and is reported to reach over 10 wt % of Asian dust particles (Mcnaughton et al., 2009). It is generally accepted that CO${}_{\mathrm{3}}^{\mathrm{2}-}$ affects atmospheric chemistry and aerosol characteristics mainly through its intrinsic alkalinity, which buffers aerosol acidity and increases SO₂ adsorption and corresponding sulfate production in the presence of oxidants (Bao et al., 2010; Kerminen et al., 2001; Yu et al., 2018; Li et al., 2007; Al-Hosney and Grassian, 2005). In fact, either CO₂ or carbonate salt is able to produce the active CO${}_{\mathrm{3}}^{\mathrm{•}-}$ under the ambient circumstance (Graedel and Weschler, 1981; Ervens et al., 2003) and is prone to increase the oxidative capacity in the atmosphere. Our early study shows that carbonate radicals serve as an active oxidant to accelerate NO₂ oxidation over mineral dust particles (Fang et al., 2021), allowing us to consider the possibility that fast heterogeneous SO₂ oxidation can be triggered by this active intermediate as well. Nevertheless, to the best of our knowledge, no work has ever considered how and to what extent the carbonate radical influences SO₂ heterogeneous oxidation in the atmosphere.

In the current study, through laboratory studies, we present that carbon dioxide and calcium carbonate, working as the precursor of carbonate radicals, have the ability to accelerate sulfate formation over authentic particles in the atmosphere. Together with quantum chemistry calculations, a detailed molecular mechanism regarding a single electron transfer (SET) process between carbonate radical and sulfite ions is elucidated. Furthermore, ground-based observations validate some findings from the laboratory-based simulations.

2 Experimental methods

2.1 Laboratory studies

2.1.1 Methodology for uptake coefficient estimation

The reaction uptake coefficient was estimated by the following Eqs. (1)–(3), as suggested by the previous work (Kong et al., 2014):

$$\begin{array}{}\text{(1)}& {\displaystyle }\mathit{\gamma }={\displaystyle \frac{\mathrm{d}\left[{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}\right]/\mathrm{d}t}{Z}}\text{(2)}& {\displaystyle }Z={\displaystyle \frac{\mathrm{1}}{\mathrm{4}}}\times A_{\mathrm{s}}\times \left[{\mathrm{SO}}_{\mathrm{2}}\right]\times v\text{(3)}& {\displaystyle }v=\sqrt{\mathrm{8}RT/\mathit{\pi }{M}_{{\mathrm{SO}}_{\mathrm{2}}}},\end{array}$$

where ν is the mean molecular velocity of SO₂, A_s is the effective sample surface, R is the gas constant, T is the temperature, $M_{{\mathrm{SO}}_{\mathrm{2}}}$ is the molecular weight of SO₂, and Z is a total number of surface collisions per unit of time. To be precise, the formation rates ($\mathrm{d}\left[{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}\right]/\mathrm{d}t$) in the equation were determined by ion chromatography (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 × 10¹⁵ ions per integrated absorption unit was obtained, and the corresponding SO₂ uptake coefficients in the TiO₂ + (CO₂) + SO₂ system were thus estimated using seven time points during the heterogeneous reaction. 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 purpose, we mixed the known proportion of K₂SO₄ and dust particles of concern and thus obtained relationship curves between the integrated absorbance of DRIFTS (i.e., diffuse reflectance Fourier transformed infrared spectroscopy) sulfate bands and corresponding theoretical sulfate content through linear fitting (Fig. S4). In total, 10 time points during the heterogeneous reaction were applied for these kinetic calculations.

2.1.2 Preparation for clay and dust membranes and investigation of sulfate formation on those authentic particles during the daytime and nighttime

Each particle suspension for Arizona test dust (ATD; 2.5 mg per 0.5 mL), IMt-2 (Illite, Montana, USA; 10 mg per 0.5 mL) and K-GA (Kaolin, Georgia, USA; 10 mg per 0.5 mL) was first dispersed into water through an ultrasonic bath for 5 min. After that, sample suspensions were transformed onto the cleansed round quartz films (d = 2 cm) using a pipette and subsequently sent to the infrared drying oven for 10 min to prepare the dust membranes. Once taken out from the oven, samples were quickly sealed into a desiccator and cooled down to room temperature before starting the experiments. A membrane sample was then mounted at the center of the reaction chamber (the top half is made of quartz, and the bottom half is made of Teflon). Before each set of experiments, a gas flow (dry air) of 300 mL min⁻¹ was introduced to the chamber for 5 min where a prepared membrane sample was installed. Afterward, samples were exposed to a 4.91 × 10¹⁴ molec. cm⁻³ SO₂ (+2.46 × 10¹⁸ CO₂ when necessary) $/$ N₂ + O₂ mixture in the absence and presence of irradiation (light intensity (I) of 30 mW cm⁻²) for 15 min before the sample were transferred to a beaker (scale of 10 mL), with 2 mL of 2 % vol. isopropanol leaching solution, then ions were extracted in the ultrasonic tank for 5 min using 0.22 µm PTFE membrane filter, followed by sending the sample into IC. Noting that dust and clay particles possess a considerable volume of sulfate ion in the background, we thus measured the background ions for each batch of synthesized dust particle membranes following the procedures described above. All data demonstrated in Fig. 6 were obtained after the subtraction of background ions.

2.1.3 Determination of gas-phase reactive oxygen species (ROS) production in the flow cell reactor

To measure the concentration of ROS released from TiO₂ particles in various reaction systems, an experimental approach using the probe molecule aniline was applied in this study. This is because compound aniline is reported to react rapidly with •OH radicals and CO${}_{\mathrm{3}}^{\mathrm{•}-}$ radicals, which are also evidenced to be two major active ROS species that are responsible for the SO₂ oxidation over mineral dust particles. The method applied in this study was almost implemented in the same way as that of the previous study (Behrman, 2018), with a slight modification. Briefly, the degradation rate of aniline in various reaction systems was monitored through high-performance liquid chromatography (HPLC; LC-10AD, Shimadzu, Japan). A Zorbax StableBond (SB) C18 (4.6 mm × 150 mm, 5 µm) reverse-phase column at 25^∘ was used with a UV detector at 236 nm to measure the aniline concentration. The mobile phase consisted of an acetonitrile/water ratio of 55 : 45 ($V/V$), with a flow rate of 1 mL min⁻¹.

The TiO₂ suspension (5 mg TiO₂ per 100 µL deionized water) was deposited onto the glass substrate (0.13–0.17 mm in thickness) using a pipette and then dried in an oven for 10 min to obtain a TiO₂-coated film. Dilute aniline solution, using a 67 mM phosphate buffer solution (pH of 7.0), was prepared and placed below the TiO₂-coated film, with an intervening gap between the TiO₂ film and solution surface around 2 mm (Fig. 1). This short distance essentially guarantees that gaseous ROS (e.g., •OH radicals or CO${}_{\mathrm{3}}^{\mathrm{•}-}$ radicals) will diffuse and react with aniline molecule (Rodriguez et al., 2013).

Figure 1 The schematic of the flow cell reaction chamber for gaseous ROS determination.

In the reaction system containing a TiO₂ film upon irradiation (the UV wavelength of 310 nm) in the presence of humidified air (RH of 30 %), when operated in a continuous mode, the overall degradation rate of the probe molecules in the presence of the TiO₂ film can be described by Eq. (4), as follows (Wang et al., 2004):

$$\begin{array}{}\text{(4)}& \begin{array}{rl}{k}_{\mathrm{obs}}& ={\displaystyle \frac{\mathrm{d}\left[\mathrm{An}\right]}{\mathrm{d}t}}=r_{\mathrm{A}}+r_{\mathrm{U}}+r_{\mathrm{ROS}}\\ & =r_{\mathrm{A}}+r_{\mathrm{U}}+k_{\mathrm{ROS},\mathrm{AN}}\left[\mathrm{ROS}\right]\left[\mathrm{An}\right],\end{array}\end{array}$$

where k_obs is the observed degradation rate of aniline, [An] is the concentration of aniline, denoted as [An] hereafter, and r_A, r_U, and r_ROS stand for aniline removal rates resulting from air stripping, UV photolysis, and ROS oxidation. k_ROS, An are the overall second-order reaction rate constants for An with ROS.

Reference experiments without the introduction of ROS were also conducted to measure r_A+r_U in each reaction system. Upon irradiation, the dust proxy TiO₂ produces hole–electron pairs, further forming •OH radicals and superoxide radicals (O${}_{\mathrm{2}}^{\mathrm{•}-}$) in the presence of absorbed water and oxygen molecules. Thus, an experiment using N₂ was adopted to investigate the role of O${}_{\mathrm{2}}^{\mathrm{•}-}$ in consuming aniline. As illustrated in Fig. 7c, a slight change in the degradation rate of aniline after stripping oxygen from the air, indicating that O${}_{\mathrm{2}}^{\mathrm{•}-}$ shows quite a smaller contribution than •OH. This result agrees well with the finding reported by Durán et al. (2019), where the removal of O${}_{\mathrm{2}}^{\mathrm{•}-}$ by adding benzoquinone (BQ) into TiO₂ suspension results in a negligible change in the An degradation rate.

Taken above, •OH radicals are assumed to be the only active ROS that accounts for the An degradation. Hence, the maximum steady concentration of •OH radicals can be given by the following equation:

$$\begin{array}{}\text{(5)}& -{\displaystyle \frac{\mathrm{d}\left[\mathrm{An}\right]}{\mathrm{d}t}}=k_{\exp }\left[\mathrm{An}\right]=k_{\mathrm{•}\mathrm{OH},\mathrm{An}}\left[\mathrm{•}\mathrm{OH}{]}_{\text{ss-max}}\right[\mathrm{An}].\end{array}$$

The integration of Eq. (5) yields the following:

$$\begin{array}{}\text{(6)}& {\displaystyle }\ln {\displaystyle \frac{[\mathrm{An}{]}_t}{[\mathrm{An}{]}_{\mathrm{0}}}}=k_{\exp }t\text{(7)}& {\displaystyle }{k}_{\exp }=k_{\mathrm{•}\mathrm{OH},\mathrm{An}}[\mathrm{•}\mathrm{OH}{]}_{\text{ss-max}}.\end{array}$$

Together with the reported second-order rate constant (k_•OH, An = 6.5 × 10⁹ M⁻¹ s⁻¹) (Samuni et al., 2002), the steady-state OH radical concentration [•OH]_ss-max in a buffered An solution can be calculated from Eq. (7). The observed degradation rate constant of k_exp can be obtained from the slope of the semi-log plot of an An degradation, as illustrated in Eq. (6). The maximum steady-state aqueous concentration of •OH supplied by the partitioning process from gas-phase •OH was thus estimated to be 2.15 × 10⁻¹⁵ M for the TiO₂ + Air system.

When CO₂ (400 ppm, parts per million; atmospherically relevant concentration) is introduced into a flow cell chamber, an increased degradation rate of An is seen, which is very likely to be the generation of active carbonate radical ions (Fig. 9). Similar to the method we adopted for the estimation of [•OH]_ss-max, reference experiments were conducted to determine the rates for air stripping and UV photolysis processes in the TiO₂ + Air+CO₂ system. In the next step, we quenched the hydroxyl radicals by adding tertiary butanol (TBA). This is because it reacts rapidly with hydroxyl radicals (Li et al., 2020) k_•OH,TBA = (6 × 10⁸ M⁻¹ s⁻¹), while showing a rather low reaction rate with carbonate radicals (Liu et al., 2015) ($k_{{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-},\mathrm{TBA}}$ < 1.6 × 10² M⁻¹ s⁻¹). Subsequently, we determined $[{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}{]}_{\mathrm{ss}}$ using the previous protocol (Huang and Mabury, 2000) with known $k_{{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-},\mathrm{AN}}$ (5.4 × 10⁸ M⁻¹ s⁻¹) (Wojnarovits et al., 2020). In the extreme case, assuming that all hydroxyl radical ions were fully trapped by absorbed and dissolved HCO${}_{\mathrm{3}}^{-}$ $/$ CO${}_{\mathrm{3}}^{\mathrm{2}-}$, the maximum steady-state CO${}_{\mathrm{3}}^{\mathrm{•}-}$ concentration was determined to be 1.39 × 10⁻¹³ M for the TiO₂ + Air+CO₂ system, matching well with an earlier study where the concentration of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ is reported to be 2 orders of magnitudes more than •OH over the water surface (Sulzberger et al., 1997).

2.2 Quantum chemical calculation

We employed density functional theory (DFT) calculations in the term of the single electron transfer (SET) process using the Gaussian 09 package to investigate this novel route (detailed in Texts S11 and S20 in the Supplement).

2.3 Field observations

2.3.1 Sampling

Our sampling for atmospheric particle matter was launched on the roof of the Department of Environmental Science and Engineering, Fudan University (Jiangwan Campus, 31.340661^∘ N, 121.506747^∘ E, 16 km away from the city downtown). More geographical information for sampling has been described in detail elsewhere (Y. Liu et al., 2020). Observations for the water-soluble ionic components of particulate matter were performed using an eight-stage non-viable cascade-impactor-type sampler (Tisch Environmental Inc., USA), the size gradients of which are in the sequence of < 0.43, 0.43–0.65, 0.65–1.1, 1.1–2.1, 2.1–.3, 3.3–4.7, 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 is conducted for the initial four stages (∼ 3.3 µm). Atmospheric airflow from the head was maintained at the constant rate of 28.3 L min⁻¹ to meet the operation criterion required for the Anderson-type sampler. Quartz filters (81 mm in diameter, Whatman, GE Healthcare, UK) were applied for the sampling, and the membranes were rinsed with ultrapure water (electrical resistivity of 18.2 MΩ) no fewer than three times; that is, they were kept in the ultrasonic cleaning tank for 40 min, then rinsed with ultrapure water twice, before being sent into the infrared drying oven, followed by packing them in aluminum foil prior to the 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, following the same aforementioned pretreatments, we measured the background concentrations of ions in blank membranes. We separated our daily samplings into two periods of 11 h each to give an insight into the influence of potential photoinduced reactions on secondary sulfate formation in the atmosphere.

2.3.2 (Bi)carbonate estimation

The concentration of bi(carbonate) ions were estimated following the protocol reported in the early works (Zhang et al., 2011; Palmer and Cherry, 1984; Fang et al., 2021; Y. Liu et al., 2020). The following two assumptions were made to simplify the estimation for bi(carbonate) concentrations within the system: (a) dominating cations in each system are H⁺, Li⁺, Na⁺, NH${}_{\mathrm{4}}^{+}$, K⁺, Mg²⁺, and Ca²⁺, whereas those transition and heavy metal ions were out of consideration, considering their limited contents of dissolve cations in the atmospheric particulate matter. (b) There were three typical organic acid ions (CH₃COO⁻, COOH⁻, and C₂O${}_{\mathrm{4}}^{\mathrm{2}-}$, which are major soluble organic acid ions in the atmosphere) taken into account for the ionization balance and the rest of the charge gap in each system was assumed to originate from (bi)carbonate ions. Then we established balance equations (Eqs. 8–11) 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).

$$\begin{array}{}\text{(8)}& {\displaystyle }\begin{array}{rl}\left[{\mathrm{H}}^{+}\right]& +\left[{\mathrm{Li}}^{+}\right]+\left[{\mathrm{Na}}^{+}\right]+\left[{\mathrm{NH}}_{\mathrm{4}}^{+}\right]+\left[{\mathrm{K}}^{+}\right]+\mathrm{2}\left[{\mathrm{Mg}}^{\mathrm{2}+}\right]\\ & +\mathrm{2}\left[{\mathrm{Ca}}^{\mathrm{2}+}\right]\\ =& \left[{\mathrm{OH}}^{-}\right]+\left[{\mathrm{F}}^{-}\right]+\left[{\mathrm{CH}}_{\mathrm{3}}{\mathrm{COO}}^{-}\right]+\left[{\mathrm{COOH}}^{-}\right]\\ & +\left[{\mathrm{NO}}_{\mathrm{2}}^{-}\right]+\left[{\mathrm{Cl}}^{-}\right]+\left[{\mathrm{NO}}_{\mathrm{3}}^{-}\right]+\mathrm{3}\left[{\mathrm{PO}}_{\mathrm{4}}^{\mathrm{3}-}\right]\\ & +\mathrm{2}\left[{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}\right]+\mathrm{2}\left[{\mathrm{C}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{4}}^{\mathrm{2}-}\right]+\left[{\mathrm{HCO}}_{\mathrm{3}}^{-}\right]+\mathrm{2}\left[{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}\right]\end{array}\text{(9)}& {\displaystyle }\left[{\mathrm{HCO}}_{\mathrm{3}}^{-}\right]={\displaystyle \frac{K_{\mathrm{1}}\left[{\mathrm{H}}_{\mathrm{2}}{\mathrm{CO}}_{\mathrm{3}}\right]}{\left[{\mathrm{H}}^{+}\right]}}\text{(10)}& {\displaystyle }\left[{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}\right]={\displaystyle \frac{K_{\mathrm{1}}{K}_{\mathrm{2}}\left[{\mathrm{H}}_{\mathrm{2}}{\mathrm{CO}}_{\mathrm{3}}\right]}{{\left[{\mathrm{H}}^{+}\right]}^{\mathrm{2}}}}.\end{array}$$

It is worth mentioning that [X] is referring to the charge concentration for ions (coulomb M). Additionally, the temperature factor was also considered to correct the equilibrium constant for (bi)carbonate ions using the following equation:

$$\begin{array}{}\text{(11)}& \mathrm{Ln}{\displaystyle \frac{K_{x^{\prime }}}{K_x}}=-{\displaystyle \frac{\mathrm{\Delta }H}{R}}\left({\displaystyle \frac{\mathrm{1}}{T_{x^{\prime }}}}{\displaystyle \frac{\mathrm{1}}{T_x}}\right),\end{array}$$

where ΔH is the temperature variation (K), R is the ideal gas constant (8.31451 J mol⁻¹ K⁻¹), and T is the temperature (K) during pH measurements. We then solved those equations to obtain a series of $\left[{\mathrm{HCO}}_{\mathrm{3}}^{-}\right]$, which were eventually corrected by subtracting blank values.

2.4 Other measurements and analysis

In addition to the above descriptions of measurements and calculation methodologies, more information, including the heterogenous reaction setup, kinetics reaction order determination, pretreatment of mineral dust, DRIFTS, IC, Raman measurements and analysis, etc., is available in Texts S1–S13 in the Supplement.

3 Results and discussion

3.1 Accelerated sulfate production in the presence of carbonate

The physicochemical properties of the employed mineral dust proxies, including Brunauer–Emmett–Teller (BET) surface area, crystal phase, and structure, were first characterized (Fig. S1), consistent with early studies (Su et al., 2008; Balachandran and Eror, 1982; Shang et al., 2010), with an additional discussion in Text S14. The spectral irradiance of the solar simulator applied in the present study is well covered by natural sunlight (Fig. S2), and we tried as much as possible to have the experimental results from the lab simulate the real atmosphere. Upon solar irradiation under RH of 30 % SO₂ $/$ N₂ + O₂ flow ([SO₂] = 2.21 × 10¹⁴ molec. cm⁻³), the sulfate production on TiO₂ + CaCO₃ mixture particles (50 wt % TiO₂ and 50 wt % CaCO₃), measured by IC, is significantly enhanced by 7 times and 23 times compared to that of pristine TiO₂ and CaCO₃ (Fig. 2a), respectively. In stark contrast, there is a negligible increase in sulfate production detected in the TiO₂ + CaCO₃ mixture relative to that of pristine CaCO₃ and TiO₂ in dark experiments (Fig. S3). Great discrepancies in sulfate production over TiO₂ + CaCO₃ particles between dark and light experiments suggest that carbonate salt may play a different role in these two scenarios. However, the alkalinity of carbonate salt favors SO₂ adsorption (Al-Hosney and Grassian, 2005; Yu et al., 2018), and the photo-oxidation process assisted by TiO₂ particles is able to strengthen the oxidation efficiency of adsorbed SO₂ (Shang et al., 2010; Chen et al., 2012), which is a plausible explanation for the increased sulfate production over TiO₂ + CaCO₃ particles. Following this speculation, two types of mixtures, TiO₂ + CaCO₃ and TiO₂ + CaO, were employed. In the dark experiments (Fig. 2b), both TiO₂ + CaO and TiO₂ + CaCO₃ almost yield an identical concentration of sulfite and sulfate. On the contrary, TiO₂ + CaCO₃ particles produce nearly 2 times more sulfate than TiO₂ + CaO particles once irradiated, along with a sharp decrease in S(IV) species on the surface of TiO₂ + CaCO₃ surfaces. Besides, CaCO₃ tends to show relatively humble physical properties, including BET surface area, surface pH, and solubility, etc., relative to CaO (see the detailed discussion in Text S15). The above results allow us to assert that the carbonate-containing system contains an alternative important mechanism for sulfate formation beyond the production of an alkaline environment (additional discussion available in Text S16). Figure 2c and d illustrate that the DRIFTS features of S(IV) and S(VI) species (Nanayakkara et al., 2014; Wu et al., 2011) increase over time on theoretical and experimental TiO₂ + CaCO₃ mixtures (wt $/$ wt = 50 $/$ 50) upon irradiation. The theoretical mixture is calculated based on the DRIFTS experiments of pristine TiO₂ and CaCO₃ particles through a simple linear superposition, whereas the experimental mixture is directly derived from the DRIFTS experiment of TiO₂ + CaCO₃ (wt $/$ wt = 50 $/$ 50) particles. These results suggest a synergistic effect presented in this mixture for sulfate formation under solar irradiation.

Figure 2 (a) Sulfate or sulfite concentration quantified by IC on TiO₂, CaCO₃, and TiO₂ + CaCO₃ particles (wt $/$ wt = 50 $/$ 50) after exposure to gaseous SO₂ under irradiation for 60 min. (b) Sulfate and sulfite concentration quantified by IC on mineral dust particles of concern after exposure to SO₂ under dark or irradiation for 20 min. In situ DRIFTS of S(IV) and S(VI) production on theoretical (c) and (d) experimental TiO₂ + CaCO₃ mixtures (wt $/$ wt = 50 $/$ 50) upon irradiation for 90 min. All spectra were processed by the Kubelka–Munk (K–M) algorithm. Note that the production of sulfur species in theoretical TiO₂ + CaCO₃ mixtures refers to 0.5 × K–M bands of sulfur species of TiO₂₋ + 0.5 × K–M bands of sulfur species of CaCO₃, while those for experimental TiO₂ + CaCO₃ mixtures refer to 1 × K–M bands of sulfur species of TiO₂ + CaCO₃ mixtures (wt $/$ wt = 50 $/$ 50). Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², total flow rate of 52.5 mL min⁻¹, and SO₂ of 2.21 × 10¹⁴ molec. cm⁻³.

Combining DRIFTS experiments with the obtained calibration curve (Fig. S4), we estimated that the uptake coefficient of TiO₂ + CaCO₃ mixture (50 wt % CaCO₃) increased by about a factor of 17 as compared to that of pure CaCO₃ or TiO₂ (Table S1). More importantly, upon irradiation, SO₂ uptake coefficients for these dust mixtures lie at the 10⁻⁴ order of magnitude, which is proven to gain importance in overall sulfate production by numerical modeling investigations (Wang et al., 2014; Zhang and Carmichael, 1999). Hence, the photochemical pathway associated with carbonate species is likely a potential driving force to increase sulfate production in the atmosphere. Meanwhile, the reaction order of SO₂ in the TiO₂ + CaCO₃ + SO₂ reaction system in the range of 400–20 000 ppb (parts per billion) is determined to be 0.80 (Fig. S5), indicating that the uptake coefficients obtained at the ppm level of SO₂ would somewhat overestimate the real one obtained at atmosphere-relevant SO₂ concentration level (several or a few tens of ppb). While we note that the difference between the lab and atmospheric conditions regarding the SO₂ concentration remains even after considering the 400 ppb case, employing hundreds of ppb SO₂ in the laboratory simulation to obtain the kinetic parameter of sulfate formation is acceptable (T. Y. Liu et al., 2020; Liu and Abbatt, 2021). Therefore, we tentatively believe that uptake coefficients estimated in this work would be valid after being calibrated.

Figure 3 (a) Energy dispersive spectroscopy (EDS) mapping of sulfur. (b) A selected HRTEM (high-resolution transmission electron microscopy) region containing a high density of sulfur for further observation is shown, and the red rectangle refers to the region shown in panel (c). (c) The HRTEM image in high resolution with lattice fringes and (d) corresponding fast Fourier transform (FFT) power spectra, lattice indexing, and (1–6) the inverse FFT analysis of lattice signal is shown in panel (d). In panel (c), the term C-SO${}_{\mathrm{4}}^{\mathrm{2}-}$ stands for crystalline SO${}_{\mathrm{4}}^{\mathrm{2}-}$, i.e., CaSO₄ and Ti(SO₄)₂. Particles employed for the HRTEM measurement refer to TiO₂ (50 wt %) + CaCO₃ (50 wt %) mixture particles upon exposure to the 4.42 × 10¹⁴ molec. cm⁻³ SO₂ $/$ N₂ + O₂ for 60 min, while other reaction conditions are as same as those of the above sulfate quantification experiments in Fig. 1. Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², and total flow rate of 52.5 mL min⁻¹.

A high-resolution transmission electron microscopy (HRTEM) analysis of TiO₂ (50 wt %) + CaCO₃ (50 wt %) particles after reaction, in combination with energy dispersive spectrometer mapping measurements of the sulfur component, was conducted to investigate the synergistic effect between TiO₂ and carbonate ions (Figs. 3a–d and S6). A region with a relatively high density of sulfur species was selected for further observation (Fig. 3a and b), and the distribution of each component was determined by fast Fourier transformation (FFT) and inverse FFT analyses (Fig. 3d) of the selected HRTEM image in high resolution with lattice fringes, as shown in Fig. 3c. The observation of crystalline Ti(SO₄)₂ and CaSO₄ on the interface of TiO₂ and CaCO₃ components implies that the synergistic effect on sulfate production likely originates from interplays of those two types of components under solar illumination.

We further assessed the importance of interfacial contact between TiO₂ and CaCO₃ in sulfate production by two synthetic approaches in which the interface abundance is modulated for comparison. Typically, a grinding method was used to make a TiO₂ + CaCO₃ mixture with compact contact between those two components, thus leading to abundant interfaces. Meanwhile, the shaking method is designed to create a TiO₂ + CaCO₃ mixture of loose contact, leaving relatively lower numbers of interfaces within the mixtures. The resulting mixing statuses of the two samples meet our expectations, evidenced by the scanning electron microscope (SEM) technique (Fig. S7). IC quantification analysis suggests that particles with considerable junctions exhibit a more pronounced promotion for sulfate production than those having relatively few junctions (Fig. S8). These results emphasize the importance of an indispensable interface contact between TiO₂ and CaCO₃ in a quick production upon irradiation.

Table 1 Chemical compositions of mineral dust simulants.

^a In the present study, CaCO₃ was taken as representative of alkaline earth metal oxide in our proxies for the authentic dust. ^b The mass ratio of four components (if any) in the simulants were controlled in the ratio of SiO₂ : Al₂O₃ : CaCO₃ : TiO₂ = 81.59 : 9.62 : 7.73 : 1.06. For instance, the mass ratio of SiO₂ to Al₂O₃ is 81.59 : 9.62 in the SiO₂ + Al₂O₃ simulant, whereas it is 81.59 : 9.62 : 7.73 in the SiO₂ + Al₂O₃ + CaCO₃ mixture. The ratios of each component are derived from the EDS mapping analysis of ATD dust particles. ^c This column refers to the molar ratio.

Figure 4 Sulfate concentration quantified by IC. Sulfate concentration was measured by IC on mineral dust simulants after exposure to gaseous SO₂ (2.46 × 10¹⁴ molec. cm⁻³) under irradiation. Noting that SiO₂ : Al₂O₃ : CaCO₃ : TiO₂ refers to the mass fraction ratios of the components in simulants. Experiments were all conducted at an RH of 30 % and light intensity (I) of 30 mW cm⁻².

The increased sulfate production was further probed by employing mineral dust simulants where two dominant crust constituents SiO₂ and Al₂O₃ were introduced into TiO₂ + CaCO₃ particles to mimic the authentic mineral dust particles in the atmosphere, with the specific component and corresponding ratio information shown in Table 1. It is worth mentioning that the determination of the ratio of each component in the simulants relies upon the EDS mapping results of ATD particles. In Fig. 4, the introduction of TiO₂ components (1 wt %) into SiO₂ + Al₂O₃ leads to 81.6 % enhancement of sulfate production because of photolabile ROS. On the other hand, a mere 24.8 wt % increase in the sulfate yield was observed once ∼ 8 wt % of CaCO₃ was incorporated into SiO₂ + Al₂O₃ dust particles. This can be attributed to the alkaline environment created by CaCO₃, which is thought to increase SO₂ adsorption (Al-Hosney and Grassian, 2005) and sulfate production accordingly. Surprisingly, mixing a ∼ 1 % mass fraction of TiO₂ and ∼ 8 wt % of CaCO₃ into SiO₂ + Al₂O₃ gives rise to a 235 % increase in sulfate formation relative to that of SiO₂ + Al₂O₃. It represents nearly an extra 100 % enhancement of sulfate production due to the presence of TiO₂ and CaCO₃ of an atmospherically relevant mass fraction. These results lead to the hypothesis that the observed synergistic effect on the heterogeneous oxidation of SO₂ is likely to take effect in the atmosphere.

Fe₂O₃ is also one of the crucial components found in authentic mineral dust (El Zein et al., 2013), and it has been reported to produce electron–hole pairs under solar irradiation (Li et al., 2019), thus likely involving the reaction channel considered in this work (see a detailed discussion in a later section). Similar to the protocol applied for synthesizing the TiO₂ + CaCO₃ mixture, αFe₂O₃ + CaCO₃ are prepared accordingly. In Fig. S9a, our results show that αFe₂O₃ cannot trigger fast SO₂ oxidation in the presence of carbonate ions upon irradiation, which is distinguished from the results we derived from the TiO₂ + CaCO₃ mixture. This can be explained by the fact that Fe₂O₃ shows a lower redox activity relative to TiO₂ (Fig. S9b), where its strong redox capability essentially enables photoinduced electrons and holes to produce O${}_{\mathrm{2}}^{\mathrm{•}-}$ and •OH radical ions. In stark contrast, the valence band and conduct band of Fe₂O₃ lie at −0.18 and at 1.68 V vs. NHE (pH = 7), lower than the redox potential required for generating an O${}_{\mathrm{2}}^{\mathrm{•}-}$, •OH and carbonate-containing ROS (Li et al., 2016). Hence, no promoted sulfate production is seen for αFe₂O₃ + CaCO₃ particles under irradiation. We also note the inconsistency between our study and the previous literature with regard to the response of SO₂ oxidation to solar irradiation over αFe₂O₃ particles, which has been interpreted in Text S17.

Overall, we show that, upon irradiation, the atmospherically relevant content of TiO₂ (nearly 1 %) in mineral dust simulants is able to interact with carbonate ions to launch an increased sulfate production, which is beyond the conventional regime of alkaline neutralization of H₂SO₄. Unlike TiO₂, αFe₂O₃ lacks the ability to initiate fast SO₂ oxidation by generating carbonate-containing ROS due to its limited photochemical activity, although ferric chemistry is important in secondary sulfate formation in the atmosphere (Sullivan et al., 2007; Yermakov and Purmal, 2003).

3.2 Accelerated sulfate production in the presence of CO₂

Atmospheric CO₂ is also an important source of (bi)carbonate. Its influence on photochemical SO₂ uptake on mineral dust was thus studied. Distinguishing S(VI) from S(IV) species over TiO₂ particles relies upon the position of the infrared (IR) bands, according to the assignment of the previous literature (Nanayakkara et al., 2014), and S(VI) and S(IV) build up as heterogeneous reactions occur in all cases (Fig. 5). In the presence of atmospherically relevant CO₂ (9.83 × 10¹⁵ molec. cm⁻³), the sulfate yield was increased under irradiation as compared to the CO₂-free case (Fig. 5a and b). We cautiously examined the net effect of formed (bi)carbonate on sulfate production by time-resolved DRIFTS spectra (Fig. 5c and d) using dark experiments. CO₂ suppresses both S(IV) and S(VI) products in the dark probably because of the competitive adsorption effect, as we observed over Al₂O₃ particles (Y. Liu et al., 2020).

Figure 5 Time-resolved DRIFTS of S(IV) and S(VI) products over TiO₂ particles after exposure to SO₂ $/$ N₂ + O₂ in the absence and presence of CO₂ upon irradiation (a, b) and those reactions under dark experiments (c, d). Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², total flow rate of 52.5 mL min⁻¹, and SO₂ of 7.37 × 10¹³ molec. cm⁻³.

The above observation leads us to speculate that the active intermediate derived from (bi)carbonate species upon irradiation is a plausible force to drive rapid sulfate formation. Besides, a nearly 50 % increase in the SO₂ uptake coefficient is observed for the mineral dust proxy TiO₂ after being exposed to the 9.83 × 10¹⁵ molec. cm⁻³ (400 ppm) CO₂ + SO₂ $/$ N₂ + O₂ mixture (Table S1). The pseudo first reaction order (1.13) was also determined in the selected concentration range of 400–20 000 ppb (Fig. S5b), which satisfies the prerequisite for the uptake coefficients derived from laboratory chambers potentially being generalized to the atmosphere condition, as we expounded in the early context.

Figure 6 Laboratory studies of sulfate production on authentic dust and clay membranes (a) K-Ga-2, (b) ATD, and (c) IMt-2 under the dark and irradiation (30 mW cm⁻²) upon exposure to 4.91 × 10¹⁴ molec. cm⁻³ SO₂ $/$ N₂ + O₂ and 2.46 × 10¹⁸ molec. cm⁻³ CO₂ + 4.91 × 10¹⁴ molec. cm⁻³ and SO₂ $/$ N₂ + O₂ at RH of 30 %. Note that the sulfate yield in three cases was normalized by the mass of dust particles employed for a heterogeneous reaction.

As another step toward a real scenario in the atmosphere, experimental trials employing authentic mineral dust particles, i.e., K-Ga-2 (Kaolin, Georgia, USA), Arizona test dust (ATD), and IMt-2 (Illite, Montana, USA) clays were implemented, with the component analysis results shown in Table 2. In Fig. 6, the pronounced increase in sulfate yield (a nearly 100 % increase in sulfate production in the CO₂-involved case under irradiation) is best seen in K-Ga-2 clay (Fig. 6a). The promotional effect of CO₂ on sulfate formation under irradiation, nonetheless, is less evident for IMt-2 (the content of TiO₂ ≈ 0.99 %) and ATD (the content of TiO₂ ≈ 0.46 %) as compared to K-Ga-2 particles. This may correlate to their higher mass fraction of alkaline earth metal oxide (denoted as AE), which enables dust particles to possess a substantial number of (bi)carbonate species within the natural environment where they have experienced long-term exposure to atmospheric CO₂ during regional transport. Therefore, the aforementioned synergetic effect takes effect over IMt-2 and ATD particles even without exposure to CO₂, presumably due to the presence of the abundant alkaline carbonate formed, and a relatively moderate increase in sulfate production was thus observed. On the other hand, TiO₂ content is not necessarily an accurate predictor of photoreactivity; the content and proportion of the active phase of TiO₂ in K-Ga-2 altogether contribute to a more pronounced increase in sulfate production relative to the other two clays (see the detailed discussion in Text S18).

Table 2 Chemical compositions of the ATD dust and standard clays.

^a Chemical compositions of the dust and clays were determined by XRF (X-ray fluorescence) results. Note that ^b n.d. refers to not detected and that ^c AE refers to alkaline earth metal oxide.

3.3 Reaction mechanism

The heterogeneous reaction of SO₂ on dust particles in the atmosphere is a complicated process, covering a series of reactions taking place in both homogeneous and heterogeneous ways. At a sufficiently low RH condition (normally below 10 % RH), water readily dissociates on the surface of metal oxide under ambient atmospheric conditions, where the metal oxide surface is terminated by hydroxyl groups that hydrogen bond to adsorbed water molecules (Cwiertny et al., 2008). In this case, SO₂ oxidation over dust particles is dominated by the reaction regime where the resulting hydroxyl groups react with gaseous SO₂ to form adsorbed S(IV)_ad species. Afterward, S(IV)_ad will be oxidized by oxidants in the atmosphere or photoinduced active intermediates produced from the dust surface upon irradiation. As the RH increases beyond 10 %–15 %, multilayer water coverage occurs, reaching approximately two monolayers at RH of 30 % (Mogili et al., 2006). Under these circumstances, the amount of water adsorbed onto the surface of the dust particles is believed to be sufficiently large so that it is liquid-like in the corresponding physical and chemical properties (Cwiertny et al., 2008; Peters and Ewing, 1997). In this work, heterogenous SO₂ oxidation over mineral dust proxies proceeds at the RH of 30 %, and two water layers are likely to attach to dust particles. Thus, radical ions are anticipated to play a key role in fast SO₂ oxidation over humidified mineral dust, and mechanism studies performed in the aqueous phase are persuasive to some extent.

Figure 7 (a) Single-wavelength transient absorption spectra of various aqueous solutions. (b) The removal rate of aniline after exposure to airflow under irradiation in the absence and presence of mineral dust particles for 300 s. Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², and a total flow rate of 52.5 mL min⁻¹. Note that an adequate amount of NaOH was introduced into the TiO₂ suspension system to achieve a pH environment condition identical to that of the TiO₂ + Na₂CO₃ suspension system. (c) Determination of sulfite and sulfate concentration after exposure to air flow under irradiation in the absence and presence of mineral dust particles for 20 min. Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², and a total flow rate of 52.5 mL min⁻¹. (d) Sulfate formation change Δ(SO${}_{\mathrm{4}}^{\mathrm{2}-}$) is determined by different sulfate concentrations with and without the addition of isopropanol as hydroxyl radical scavenger. (e) The difference in transient absorption kinetics of the sulfite radical and carbonate radical at the various aqueous solutions and their corresponding growth–decay fit curves. A ΔA signal was recorded at 255 and 600 nm after a pulsed 355 nm laser excitation. (f) Electron spin resonance (ESR) spectrometry of [DMPO–SO${}_{\mathrm{3}}^{\mathrm{•}-}$] intermediary formed in a solution of TiO₂ (3 mg per 4 mL) + 0.1 M Na₂SO₃ and TiO₂ (3 mg per 4 mL) + 0.5 M Na₂CO₃ + 0.1 M Na₂SO₃. For clarity, the integrated areas of the ESR profiles were also presented for direct comparison. Note that Exp. and Sti. stand for experimental results and corresponding fitting results using isotropic radicals software.

Our preliminary sulfate quantification results (Fig. 2a and b) suggest that the presence of (bi)carbonate ions under solar light contributes to increased sulfate yield. In this carbonate-containing reaction system, a plausible intermediate is the active CO${}_{\mathrm{3}}^{\mathrm{•}-}$. It is readily produced via the following two pathways. First of all, a carbonate anion can be directly oxidized by photoinduced holes produced from typical mineral dust upon solar irradiation (Reactions R1 and R2), as the redox potential of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ $/$ CO${}_{\mathrm{3}}^{\mathrm{2}-}$ is 1.78 V (vs. NHE at pH = 7), which is lower than the TiO₂ valence band (VB) potential of 2.67 V (vs. NHE at pH = 7), as follows (Li et al., 2016; Xiong et al., 2016):

$$\begin{array}{}\text{(R1)}& {\displaystyle }\mathrm{Mineral}\mathrm{dust}+hv\to h^{+}+e^{-}\text{(R2)}& {\displaystyle }{h}^{+}+{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}.\end{array}$$

In the second pathway, carbonate radicals evolve through the reaction of (bi)carbonate anion with hydroxyl radicals •OH forming over mineral dust surfaces (Zhang et al., 2015; Reaction R3 and R4).

$$\begin{array}{}\text{(R3)}& {\displaystyle }{h}^{+}+{{\mathrm{H}}_{\mathrm{2}}\mathrm{O}}_{\mathrm{ad}}\to \mathrm{•}\mathrm{OH}+{\mathrm{H}}^{+}\text{(R4)}& {\displaystyle }\mathrm{•}\mathrm{OH}+{\mathrm{HCO}}_{\mathrm{3}}^{-}/{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}/{\mathrm{OH}}^{-}.\end{array}$$

The above assumptions are supported by nanosecond transient absorption spectra (NTAS), in which a signal (ΔOD) of a carbonate radical CO${}_{\mathrm{3}}^{\mathrm{•}-}$ at 600 nm (Bhattacharya et al., 1998) only emerges for dust suspension containing (bi)carbonate species (Fig. 7a). An increased degradation rate of aniline observed in TiO₂ suspension due to the presence of carbonate ions produces additional evidence of the formation of active CO${}_{\mathrm{3}}^{\mathrm{•}-}$ ions and the strengthened oxidation capability of TiO₂ (Fig. 7b; see the additional discussion in Text S19). The CO${}_{\mathrm{3}}^{\mathrm{•}-}$-induced chemistry was further evidenced by •OH scavenging experiments using tertiary butyl alcohol (TBA) and isopropanol (i-PrOH), as they show lower reaction rates with CO${}_{\mathrm{3}}^{\mathrm{•}-}$ ($k_{{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-},\text{i-PrOH}}$ < 4.0 × 10⁴ M⁻¹ s⁻¹ and $k_{{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-},\mathrm{TBA}}$ < 1.6 × 10² M⁻¹ s⁻¹) relative to those with •OH (k_{i-PrOH∼•OH} < 1.9 × 10⁹ M⁻¹ s⁻¹ and (k_•OH,TBA = 6 × 10⁸ M⁻¹ s⁻¹; Buxton et al., 2009; Liu et al., 2015; Li et al., 2020). TBA dramatically decreases the yield of sulfate on the TiO₂ surface by nearly 70 %, with sulfite ions being the dominant sulfur species (Fig. 7c). Meanwhile, a great loss of sulfate yield when TiO₂ suspension was added with i-PrOH (Fig. 7d). This is in strong contrast to the result of a carbonate-involved system, where the reactivity is sustained as carbonate radicals offer an alternative reaction pathway for SO₂ oxidation. This is plausible since the carbonate ions are an excellent •OH scavenger, and CO${}_{\mathrm{3}}^{\mathrm{•}-}$ becomes the predominant species in a relatively strong alkaline aqueous-like environment in the presence of carbonate salt. This is supported by a previous work (Sun et al., 2016) in which adding 0.1 M of bicarbonate salt into the UV $/$ H₂O₂ system (H₂O₂ = 0.3 mM) was sufficient to suppress •OH concentration to around 10⁻¹⁵ M, creating a carbonate-radical-dominated regime ([CO${}_{\mathrm{3}}^{\mathrm{•}-}$] = 8.64 × 10¹² M). In our experiments (Fig. 7b), 0.2 M of carbonate salt was employed, and the reaction rate of CO${}_{\mathrm{3}}^{\mathrm{2}-}$ with •OH is nearly 2 orders of magnitude higher than that of HCO${}_{\mathrm{3}}^{-}$, thus giving rise to carbonate radical being the substitute for the hydroxyl radical in the reaction. The above results suggest that •OH is a major contributor to the sulfate yield on TiO₂ particles in the absence of carbonate ions, while CO${}_{\mathrm{3}}^{\mathrm{•}-}$ ions dominate the SO₂ oxidation over humidified carbonate-containing TiO₂ particles upon irradiation. In addition to experimental investigations, the carbonate radical formation process is proved to be thermodynamically favorable and supported by density functional theory (DFT) calculations (Fig. S10 and Text S20).

On the other hand, early studies (Das, 2001; Neta and Huie, 1985; Chameides and Davis, 1982) agree with the key role of the sulfite radical (SO${}_{\mathrm{3}}^{\mathrm{•}-}$) in rapid sulfate production in an aqueous medium, and the present reaction system creates a localized environment in which the SO${}_{\mathrm{3}}^{\mathrm{•}-}$ can be readily produced from the TiO₂ and S(IV) species upon solar illumination (Salama et al., 1995). Consequently, a probe light of NTAS at wavelength 255 nm (ascribed to the sulfite radical) and 600 nm (ascribed to carbonate radical) were simultaneously monitored (Hayon et al., 1972; Ghalei et al., 2016; Goldstein et al., 2001). A weak signal of sulfite radical was observed in the system of the TiO₂ + Na₂SO₃ suspension under irradiation (Fig. 7e). On the contrary, the sulfite radical signal is strengthened after the introduction of carbonate ions into the TiO₂ + Na₂SO₃ suspension, along with a significant decrease in the signal for carbonate radical. Electron spin resonance (ESR) data (Fig. 7f) further confirm the increase in the SO${}_{\mathrm{3}}^{\mathrm{•}-}$ after 2 min UV irradiation in the presence of a carbonate ion. Based on the above results, one may deduce that the interplay between carbonate radical and sulfite ions is a crucial step giving rise to the increased SO${}_{\mathrm{3}}^{\mathrm{•}-}$, which is reported to account for rapid atmospheric sulfate formation through chain propagation reactions that involve SO${}_{\mathrm{4}}^{\mathrm{•}-}$ and SO${}_{\mathrm{5}}^{\mathrm{•}-}$ intermediates (Hung and Hoffmann, 2015; Hung et al., 2018). Additionally, this sulfite radical ion chemistry is believed to drive fast sulfate formation over mineral dust particles as well (Gankanda et al., 2016; Rubasinghege et al., 2010). Nevertheless, there are two possibilities that might explain the aforementioned interaction. One way is through the oxygen transfer and the other route is electron transfer, which needs further clarification.

Figure 8 (a) In situ DRIFTS of heterogeneous reaction of SO₂ on the TiO₂ particles for 2 and 60 min after being exposed to C¹⁶⁽¹⁸⁾O₂ $/$ N₂ for 20 min under irradiation. (b) Reaction pathway of the interaction between the carbonate radical (CO${}_{\mathrm{3}}^{\mathrm{•}-}$) and sulfite (SO${}_{\mathrm{3}}^{\mathrm{2}-}$) and the (c) interaction between the hydroxyl radical (•OH) and sulfite (SO${}_{\mathrm{3}}^{\mathrm{2}-}$) through the SET process at the coupled cluster calculations with all single and double excitations (CCSDs) (T)-F12/cc-PVDZ-F12//M06-2X/6-311$++$G (3df, 3pd) level (where polarized valence double zeta is PVDZ), and $\mathrm{\Delta }{G}_{\mathrm{0}}^{\mathrm{SET}}$ represents the difference in the Gibbs free energy between the reactant and product. The white, black, yellow, and red spheres represent H, C, S, and O atoms, respectively. In order to visualize the variation in the surface products in oxygen isotope experiments (a), DRIFTS features of these concerned species are highlighted in dark colors.

We first examined the oxygen transfer path through ¹⁸O isotope labeling experiments. TiO₂ particles were initially exposed to C¹⁶O₂ $/$ N₂ and C¹⁸O₂ $/$ N₂, followed by the exposure of SO₂ $/$ N₂ + O₂ under irradiation (Fig. 8a). A bidentate carbonate band centered at 1573 cm⁻¹ appears after the introduction of C¹⁶O₂ $/$ N₂, while this band shifts to 1558 cm⁻¹ when C¹⁸O₂ $/$ N₂ is introduced, indicating the incorporation of ¹⁸O into bidentate carbonate species, in accordance with a previous report (Liao et al., 2002). However, no shift of IR features at 1269, 1219, and 1159 cm⁻¹, assigned to (bi)sulfate species on TiO₂ particles, was observed throughout the reaction. This implies that the oxygen transfer path does not account for the rapid SO₂ oxidation on particles of concern.

In light of the above analysis, the electron transfer might be a plausible pathway to explain the fast oxidation within the reaction system. DFT calculations provide an accessible approach to studying the electron transfer pathway. The result in Fig. 8b illustrates SO${}_{\mathrm{3}}^{\mathrm{•}-}$ formation is a SET process of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and SO${}_{\mathrm{3}}^{\mathrm{2}-}$, where the O atom in SO${}_{\mathrm{3}}^{\mathrm{2}-}$ transfers an electron to the O atom in CO${}_{\mathrm{3}}^{\mathrm{•}-}$ to form SO${}_{\mathrm{3}}^{\mathrm{•}-}$ and CO${}_{\mathrm{3}}^{\mathrm{2}-}$. This SET reaction is a thermodynamically favorable process, with the difference in the Gibbs free energy between reactant and product lying at −24.09 kcal mol⁻¹. We note that the insufficient O₂ supply in aqueous media may be an underlying constraint to the proposed CO${}_{\mathrm{3}}^{\mathrm{•}-}$-initiated SO₂ oxidation pathway. Therefore, a careful estimation of both the oxygen consumption and supply rates were conducted, revealing that oxygen supply flux can be sufficiently larger than corresponding consumption (see additional details in Text S21). This enables us to deduce that considered chain reactions can continually proceed. Taking the above results and discussions altogether, the following reactions are proposed accordingly (Reactions R5–R8):

$$\begin{array}{}\text{(R5)}& {\displaystyle }{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{CO}}_{\mathrm{3}}^{\mathrm{2}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{•}-}\text{(R6)}& {\displaystyle }{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{•}-}+{\mathrm{O}}_{\mathrm{2}}\to {\mathrm{SO}}_{\mathrm{5}}^{\mathrm{•}-}\text{(R7)}& {\displaystyle }{\mathrm{SO}}_{\mathrm{5}}^{\mathrm{•}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{SO}}_{\mathrm{4}}^{\mathrm{•}-}+{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}\text{(R8)}& {\displaystyle }{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{•}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{•}-}.\end{array}$$

Another important pathway needs to be considered as well; that is, SO${}_{\mathrm{3}}^{\mathrm{•}-}$ can also be formed via the conventional reaction of •OH and SO${}_{\mathrm{3}}^{\mathrm{2}-}$ (Reaction R9).

$$\begin{array}{}\text{(R9)}& \mathrm{•}\mathrm{OH}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\to {\mathrm{SO}}_{\mathrm{3}}^{\mathrm{•}-}+\mathrm{OH}.\end{array}$$

Figure 9 The degradation rate of aniline after exposure to air flow under irradiation in the absence (a) and presence (b) of CO₂ over mineral dust proxy particles TiO₂ as a function of the reaction time. Reaction conditions are RH of 30 %, light intensity (I) of 30 mW cm⁻², and a total flow rate of 52.5 mL min⁻¹.

In this SET process, the electron donor SO${}_{\mathrm{3}}^{\mathrm{2}-}$ reacts spontaneously with electron acceptor •OH (Fig. 8c), and the calculated activation free energy barrier $\mathrm{\Delta }{G}_{\mathrm{SET}}^{\ne }$ for this SET reaction is 2.50 kcal mol⁻¹. Hence, the reaction process of •OH with SO${}_{\mathrm{3}}^{\mathrm{2}-}$is diffusion controlled, and the total rate constant k_SET−2 was calculated to be 7.12 × 10⁹ M⁻¹ s⁻¹. In comparison, the rate constant k_SET−1 of the diffusion-controlled SET process for CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and SO${}_{\mathrm{3}}^{\mathrm{2}-}$ was estimated to be 7.42 × 10⁹ M⁻¹ s⁻¹. Despite a slight net increase in the rate, the distinguishable concentration of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and •OH should also be taken into account for the rate comparison in varied reaction paths. To visualize the difference, relative rates were calculated according to Eq. (12), as follows:

$$\begin{array}{}\text{(12)}& r={\displaystyle \frac{v_{{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}}}{v_{\mathrm{•}\mathrm{OH}+{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}}}}={\displaystyle \frac{k_{\mathrm{SET}-\mathrm{1}}\left[{\mathrm{CO}}_{\mathrm{3}}^{\mathrm{•}-}\right]\left[{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\right]}{k_{\mathrm{SET}-\mathrm{2}}\left[\mathrm{•}\mathrm{OH}\right]\left[{\mathrm{SO}}_{\mathrm{3}}^{\mathrm{2}-}\right]}},\end{array}$$

where r is the ratio of two reaction rates, and [CO${}_{\mathrm{3}}^{\mathrm{•}-}$], [SO${}_{\mathrm{3}}^{\mathrm{2}-}$], and [•OH] refer to the concentration of corresponding reactants. In the literature, it is suggested that the concentration of carbonate radicals is able to show 2 orders of magnitude higher than that of hydroxyl radicals at the surface of the water under solar irradiation (Chandrasekaran and Thomas, 1983; Goldstein et al., 2001; Sulzberger et al., 1997), which is consistent with the concentration gap between carbonate radicals and hydroxyl radicals through a partitioning process from the gas phase determined in our reaction system (Fig. 9). While the net concentrations of carbonate and hydroxyl radicals in the water layers of humified particles are very likely to be different from those found in the bulk aqueous media, the concentration inputs of two radicals with the gap of 2 orders of magnitude could somehow reflect the relative contribution of carbonate radicals and hydroxyl radicals to sulfate production, based on the literature results and our experimental trails. The concentrations of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and •OH were set in the range from 1.0 × 10⁻¹⁰ to 1 × 10⁻¹² mol L⁻¹ and from 1.0 × 10⁻¹² to 1 × 10⁻¹⁴ mol L⁻¹ (Sulzberger et al., 1997), and the r value could thus reach to 1.04 × 10⁴ at most (Fig. S11). As a result, we speculate that the formation pathway of SO${}_{\mathrm{3}}^{\mathrm{•}-}$ via an interaction between CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and SO${}_{\mathrm{3}}^{\mathrm{2}-}$ is a more efficient route that corresponds well with our experimental results.

In addition to the pathway launched by photo-generated holes, the sink of photo-generated electrons is also considered. In our reaction system, O₂ is thought to be an electron trap and to produce the superoxide radical ions (O${}_{\mathrm{2}}^{\mathrm{•}-}$), which are reported to play a non-negligible role in sulfate formation (Shang et al., 2010) and should be taken into account to give a whole picture of reaction scheme in triggering sulfate formation on the surface of TiO₂-containing mineral dust particles. p-benzoquinone is a commonly used O${}_{\mathrm{2}}^{\mathrm{•}-}$ scavenger for trapping the O${}_{\mathrm{2}}^{\mathrm{•}-}$ radical ions (Yan et al., 2018). Our data show that adding an excess amount of p-benzoquinone into TiO₂ particles reduces the sulfate yield by 32 %, along with the appearance of sulfite ions over TiO₂ particles upon exposure to SO₂ (Fig. 7c). Notably, the decrease in sulfate yield by around 30 % in the presence of O${}_{\mathrm{2}}^{\mathrm{•}-}$ scavenger p-benzoquinone is almost complementary to that added with •OH scavenger using TBA (70 %), pointing toward a minor sulfate formation pathway contributed by O${}_{\mathrm{2}}^{\mathrm{•}-}$ relative to the major pathway by CO${}_{\mathrm{3}}^{\mathrm{•}-}$ when carbonate ions are presented to efficiently capture •OH ions. Following the work of Shang et al. (2010), O${}_{\mathrm{2}}^{\mathrm{•}-}$-involved SO₂ oxidation can be given as Reactions (R10)–(R12), as follows:

$$\begin{array}{}\text{(R10)}& {\displaystyle }{e}^{-}+{\mathrm{O}}_{\mathrm{2}}\to {\mathrm{O}}_{\mathrm{2}}^{\mathrm{•}-}\text{(R11)}& {\displaystyle }{\mathrm{SO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{2}}^{\mathrm{•}-}\to {\mathrm{SO}}_{\mathrm{3}}+{\mathrm{O}}^{\mathrm{•}-}\text{(R12)}& {\displaystyle }{\mathrm{SO}}_{\mathrm{3}}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\to {\mathrm{H}}_{\mathrm{2}}{\mathrm{SO}}_{\mathrm{4}},\end{array}$$

where intermediate SO₃, formed via the interaction between SO₂ and O${}_{\mathrm{2}}^{\mathrm{•}-}$, subsequently couples with water molecules to produce sulfate species as a final product. pH is an important factor within aqueous chemical reaction processes and is preferred to alter the dominant regime for sulfate production. Yet, so far, adjusting the pH of particle surfaces is quite tough, and exploring the role of dust surface pH in the reactivity of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ is not easily achieved. Notwithstanding, the increase in pH in TiO₂ suspension was observed to promote the production of CO${}_{\mathrm{3}}^{\mathrm{•}-}$, further strengthening the oxidation capability of dust particles (Fig. 7b). In contrast, decreasing pH is expected to reduce the yield of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ since the reaction rate of CO${}_{\mathrm{3}}^{\mathrm{2}-}$ with •OH is nearly 2 orders of magnitude higher than that with HCO${}_{\mathrm{3}}^{-}$. On this basis, a question arises as to whether the surface pH of mineral dust can be sustained to maintain the fast SO₂ oxidation triggered by CO${}_{\mathrm{3}}^{\mathrm{•}-}$ in the typical lifespan of mineral dust.

Considering that the SO₂ concentration employed in this work is higher than that in the real atmosphere, the concept of equivalent exposure time is introduced to evaluate the influence of pH on the SO₂ oxidation pathway initiated by CO${}_{\mathrm{3}}^{\mathrm{•}-}$ (see a more detailed discussion on determining equivalent exposure time in Text S22). The heterogeneous sulfate production over TiO₂ and TiO₂ + CaCO₃ particles versus equivalent exposure time were plotted (Fig. S12). Clearly, the sulfate yield builds up steadily during the 2-week equivalent exposure time, suggesting that the regime of CO${}_{\mathrm{3}}^{\mathrm{•}-}$-initiated SO₂ oxidation over TiO₂ and TiO₂ + CaCO₃ particles is slightly affected by the possible decrease in surface pH because of the accumulation of sulfate production over entire reaction course. In the atmosphere, the lifetime of mineral dust particles ranges from several days to weeks (Bauer and Koch, 2005), and the equivalent exposure time considered in this study (nearly 2 weeks) falls right within the characteristic lifespan range of mineral dust particles. Besides, 20 ppb is assumed to be an atmospherically relevant concentration to calculate equivalent exposure time in this study, whereas even low SO₂ concentrations (several or a few tens ppb of SO₂) were monitored in the field observation (He et al., 2014; Watanabe et al., 2020). Therefore, the reduction of dust surfaces pH would be more moderate than what we now consider, and even a little influence of surface pH on our proposed reaction scheme would have an impact. Therefore, persistent growth of sulfate shows a negligible effect on the CO${}_{\mathrm{3}}^{\mathrm{•}-}$-initiated SO₂ oxidation scheme proposed in this work.

Additionally, dust particles are reported to eject the radical ions from the surface under solar light irradiation, severing as an underlying pathway for sulfate aerosol formation in the atmosphere (Chen et al., 2021; Dupart et al., 2012) and is described as follows:

$$\begin{array}{}\text{(R13)}& {\displaystyle }\mathrm{Mineral}\mathrm{Dust}+hv\to \mathrm{ROS}\left(\mathrm{g}\right)\text{(R14)}& {\displaystyle }\mathrm{ROS}\left(\mathrm{g}\right)+\mathrm{humidified}\mathrm{Air}+{\mathrm{SO}}_{\mathrm{2}}\to \mathrm{Sulfate}\left(\mathrm{g}\right),\end{array}$$

where ROS (g) stands for the active intermediates in the gas phase. Over 400 ppm of CO₂ is universal in the atmosphere, and it is expected to form (bi)carbonate ions once enters the quasi-liquid layer of humified particles. Bi(carbonate) ions are then prone to reacting with hydroxyl radical ions to form carbonate radicals. Following this line of reasoning, we attempt to monitor the plausible gas ROS species that are formed in the presence of CO₂ (see the detailed discussion in Sect. 2).

When CO₂ (atmospherically relevant concentration) is introduced into the homemade flow cell chamber, with the intervening gap between the TiO₂-coated film and probe molecule solution fixing at nearly 2 mm, the short distance guarantees that gaseous ROS will diffuse and react with aniline (Rodriguez et al., 2013). An increased degradation rate of this probe molecule was seen, which can be speculated to be the generation of active carbonate radical ions (Fig. 9). Aqueous CO${}_{\mathrm{3}}^{\mathrm{•}-}$ is believed to be supplied by the partitioning processes from CO${}_{\mathrm{3}}^{\mathrm{•}-}$ (g) that come from humified dust particles, and its maximum steady-state concentration was determined to be 1.39 × 10⁻¹³ M for TiO₂ + Air + CO₂ system, which is over 1.8 orders of magnitudes higher than that of •OH for the TiO₂ + Air system (2.15 × 10⁻¹⁵ M). This observation matches with an earlier study, where the concentration of a carbonate radical can be nearly 2 orders of magnitudes greater than •OH over the water surface (Sulzberger et al., 1997).

Overall, the above results suggest that the photochemistry that involves carbonate ions, more precisely CO${}_{\mathrm{3}}^{\mathrm{•}-}$ radicals, increases sulfate production. This finding broadens the prevailing view that the acceleration of SO₂ oxidation over the carbonate salt is merely due to the favorable neutralization of H₂SO₄ over an alkaline surface. Worth noting is that upon irradiation the active component TiO₂ in mineral dust produces carbonate radicals in the gas phase when CO₂ presents, therefore potentially promoting sulfate aerosol formation in the atmosphere. Overall, it could be speculated that (bi)carbonate species strengthen the oxidative capacity of TiO₂-containing dust particles with regard to SO₂ oxidation.

3.4 Field measurements of sulfate and (bi)carbonate ions

Complementary field sampling and analysis were further conducted to examine our hypothesis that intermediate CO${}_{\mathrm{3}}^{\mathrm{•}-}$ may play role in secondary sulfate formation in the atmosphere. We first considered the meteorological condition of wind speed, which is an important parameter determining whether the local chemical process gains importance in affecting secondary sulfate formation. Meteorological information was collected from an open-access database (https://www.aqistudy.cn/, last access: 7 September 2021). During the sampling period, the wind scale mainly varies from 0 to 1, corresponding to the wind speed ranging from 0 to 1.5 m s⁻¹ (Fig. S13). All plots shown in Fig. S13 give rise to a statistical wind speed of 0.76 ± 0.73, which represents the weak dispersion of pollutants at a low wind speed (not exceeding 2.5 m s⁻¹) (Witkowska et al., 2016; P. Liu et al., 2020), indicating that a local source is a dominant contributor to the air pollution.

Under stagnant meteorological conditions (wind speed < 1.5 m s⁻¹), for the coarse mode (2.5–10 µm) of sulfate, the heterogeneous reaction of SO₂ on the dust surfaces is thought to be a major contributor (Liu et al., 2017). This correlates to the fact that a large mass fraction of mineral dust is abundant in the coarse-mode particulate matter (PM; Fang et al., 2017; Miller-Schulze et al., 2015) in which TiO₂ was found at mass mixing ratios ranging from 0.1 % to 10 % depending on the exact location where particles were lifted up (Hanisch and Crowley, 2003; Chen et al., 2012). Therefore, PM with relatively larger size dimensions are expected to contribute to secondary sulfate formation via heterogeneous reactions, which is supported by the recent field study in which the carbonate fraction of coarse PM is evidenced to promote secondary sulfate production (Song et al., 2018). Considering this, rather than determining the concentration of water-soluble ions in all stages, special attention is focused on PM collected in stages 1–4 (particles with their dimension of ≥ 3.3 µm). As (bi)carbonate ions are known as key precursors in producing CO${}_{\mathrm{3}}^{\mathrm{•}-}$ and accelerating sulfate formation, quantifications of those relevant water-soluble ions were thus conducted (see details in Sect. 2 and Text S13).

Figure 10 Field observation for the relationship between carbonate and sulfate ions during daytime and nighttime hours. Linear relationship analyses for measured sulfate ions and estimated carbonate ions (a) and for measured sulfate ions and estimated bicarbonate ions (b) during the daytime and nighttime hours, with particle sizes of PM ranging from 3.3 to 9 µm.

We further consider the relationships between sulfate ions and (bi)carbonate ions by means of linear regression analysis. However, under the low wind speed (0.76 ± 0.73), correlation coefficients R² obtained for the relationship between bi(carbonate) and sulfate ions are not promising, with 0.56 (sulfate vs. carbonate) and 0.61 (sulfate vs. bicarbonate) for PM_3.3-PM_9.0 during daytime hours (Fig. 10). A plausible explanation is that in spite of having little significance, a local primary emission source also brings bias and uncertainty to the correlation results. Shanghai is a coastal city, and sulfate species such as K₂SO₄ and Na₂SO₄ from the sea salt contribute to the local sulfate emission as well (Long et al., 2014). On the other hand, this novel SO₂ oxidation channel is yet to be in the infant stage, and only active mineral dust components have been considered in this work, whereas other components found in the coarse mode of PM such as organic matter, elemental carbon, and sea salt (Cheung et al., 2011) are likely to involve this mechanism and alter the response of sulfate yield to SO₂ heterogeneous uptake. In addition, the water-soluble ions determined in these samples may not come from the net contribution of heterogenous reaction processes in absolute daytime and nighttime periods. In other words, some of the collected samples, experiencing the heterogeneous reaction that occurs during the day(night)–night(day) shifts periods, will inevitably be assigned to the sulfate ions measured in separate sampling hours, thus reducing the correlation coefficients.

For those large particles (LPs), which refer to the particles with a diameter large than 9 µm in this work, sulfate ions show a rather weak or even no correlation to (bi)carbonate ions during the nighttime or daytime hours (Fig. S14). This is likely due to the short lifetime of the LPs. Generally, the aerosol lifetime is of the order of less than an hour to days (Koelemeijer et al., 2006), highly depending on the particle size. For example, the lifetime of PM₁₀ ranges from minutes to hours, and its travel distance, in general, is less than 10 km (Agustine et al., 2018). As a consequence, secondary sulfate formation through chemical reaction over LPs is not significant with respect to in situ emissions. When PM downsizes to 2.5 µm, then PM_2.5 has a lifetime prolonged to nearly 1 d or longer (Wu et al., 2020). Therefore, PM_3.3–PM_9.0 are expected to have a relatively long lifetime, of the order of several hours on average, which enables the heterogeneous reaction process to become a more important contributor to overall sulfate measured in PM_3.3–PM_9.0 than that in PM_≥9.0. This is supported by our observations where, during the daytime hours, the correlation coefficients for PM_3.3–PM_9.0, i.e., 0.56 (sulfate vs. carbonate) and 0.61 (sulfate vs. bicarbonate), are higher than that of PM_≥9.0, i.e., 0.489 (sulfate vs. carbonate) and 0.36 (sulfate vs. bicarbonate), respectively. Likewise, higher correlation coefficients are also observed for PM_3.3–PM_9.0 than PM_≥9.0 in the sample collected during the nighttime periods.

While we note that the correlations between sulfate and (bi)carbonate are not high in this work, ground-based field measurements of sulfate and (bi)carbonate ions shed light on their distinct correlations during the daytime and nighttime hours. In Figs. 10 and S14, the negative correlations between the mass concentrations of sulfate ions and (bi)carbonate ions are observed in the nighttime hours, consistent with the suppression of sulfate formation by CO₂ in the dark experiments. This is also supported by lab-based observations where CO₂-derived (bi)carbonate species are demonstrated to suppress sulfate production over two dominant mineral dust components, i.e., aluminum oxide (Y. Liu et al., 2020) and silicon dioxide (Fig. S15 and Text S23). Alternatively, while CO₂-derived (bi)carbonate may slightly affect sulfate accumulation over PM with high water content in the dark scenario, fresh PM is usually dry when emitted into the atmosphere. Due to the competitive adsorption, the occurrence of the suppression of SO₂ adsorption and subsequent sulfate formation is possible in the early emission stage before PM becomes wet, thus contributing to the overall negative correlation.

Instead, positive correlations were seen for those ions within PM sampled during the daytime hours, regardless of the size ranges and carbonate types (HCO${}_{\mathrm{3}}^{-}$ $/$ CO${}_{\mathrm{3}}^{\mathrm{2}-}$). This matches with the scenarios in which sulfate production, upon irradiation in the presence of (bi)carbonate ions, is increased over both model and authentic dust particles. Except for this case (nighttime period, size larger than 9 µm), most of the significance P values for their correlations were smaller than 0.1, specifically with significant P values below 0.5 determined for bicarbonate vs. sulfate, implying the plausible underlying connection between sulfate and (bi)carbonate ions. In fact, preceding ground-based observations of a highly correlated relationship between Ca²⁺ and SO${}_{\mathrm{4}}^{\mathrm{2}-}$ water-soluble ions (Wu et al., 2020) during the carbonate-enriched dust storm episodes, together with persistent reports on the significant role of photochemical channels in elevating the sulfate concentration level during the daytime hours (Wei et al., 2019; Kim et al., 2018; Wu et al., 2017), indirectly reflects the possibility of accelerated SO₂ oxidation triggered by photo-generated active intermediates associated with carbonate species.

Overall, this is the first time that relationships between those ions have been explored separately in these two periods. Taken together, carbonate radicals are likely to promote sulfate production in the atmosphere during daytime hours. A detailed and systematic SO₂ oxidation channel triggered by CO${}_{\mathrm{3}}^{\mathrm{•}-}$ needs further investigation to enable a better interpretation of the correlations between these inorganic ions at the given meteorological conditions of sampling and physicochemical properties of PM.

4 Conclusion

On the basis of the experimental and theoretical results derived from this work, we, for the first time, propose a novel reaction channel for fast SO₂ oxidation over mineral dust particles due to the formation of carbonate radical ions. A schematic chart for the sulfate formation in the presence of carbonate radicals upon solar light or bi(carbonate) ions under dark conditions is summarized and elucidated in Fig. 11. During the nighttime hours at 298 K (ambient temperature) CO₂-derived (bi)carbonate species are prone to having a slightly negative effect on sulfate formation, presumably due to the competitive adsorption between CO₂ and SO₂. For alkaline carbonate salt, it favors sulfate formation through the neutralization process. On the other hand, in the daytime, both CO₂-derived (bi)carbonate species and carbonate salt work as the precursors of CO${}_{\mathrm{3}}^{\mathrm{•}-}$, which promotes sulfate formation. In particular, uptake coefficients for carbonate salt containing mineral dust can be increased 17 times, which is more pronounced than the increase due to the neutralization regime in the dark condition. Consistent with the findings reported in the early studies (Chen et al., 2021; Dupart et al., 2012), we speculate the production of gas-phase CO${}_{\mathrm{3}}^{\mathrm{•}}$ ions when mineral dust particles are irradiated in the presence of CO₂ (atmospherically relevant concentration 400 ppm). This observation potentially implies that the increased sulfate yield in part comes from increased external secondary sulfate aerosol triggered by CO${}_{\mathrm{3}}^{\mathrm{•}-}$ (g).

Figure 11 Schematic of the sulfate formation in the presence and absence of a carbonate radical. Note that g and ad represent gas phase and adsorbed carbonate radical ions, respectively.

By means of ROS scavenger experiments, direct observation of carbonate radicals using NTAS analysis, oxygen isotope assay, ESR measurement, and DFT calculations, CO${}_{\mathrm{3}}^{\mathrm{•}-}$-initiated S(IV) oxidation involving single electron transfer process is elucidated. While carbonate radical ions are mainly responsible for rapid sulfate formation, superoxide radical ions are likely to serve as a minor pathway over TiO₂-containing mineral dust particles (not shown in the schematic chart to direct readers' focus on CO${}_{\mathrm{3}}^{\mathrm{•}-}$). In addition, a weak correlation between sulfate ions and (bi)carbonate ions observed for PM_3.3–PM_9.0 in this work reasonably correlates to non-chemical primary emission and the complicated nature of CO${}_{\mathrm{3}}^{\mathrm{•}-}$ regime of sulfate production in the atmosphere. Nonetheless, complementary field sampling of ambient PM and analysis of sulfate and (bi)carbonate ions in this study unfold their distinct correlations during the daytime and nighttime hours; these two tendencies agree with the experimental observations.

In this work, only atmospheric secondary sulfate formation was considered, whereas the oxidation of primary organic species has not yet been investigated. In fact, carbonate radical ions are prone to reacting rapidly with electron-rich organic amines (Stenman et al., 2003; Yan et al., 2019) and phenol (Xiong et al., 2016; Busset et al., 2007), and they may potentially serve as the key oxidants that drive the fast formation of secondary organic aerosol in the atmosphere. Besides, an observation of the strengthened photochemistry launched by carbonate radicals suggests that such chemistry may be amplified on atmospherically relevant reactions that occur in cloud droplets and fog water, where they often contain hydroxyl radicals and water-soluble (bi)carbonate ions.

Importantly, gas-phase carbonate radical ions are speculated to be formed in the atmospherically relevant CO₂ concentration (400 ppm) when mineral dust is irradiated. This will help in the formation of external sulfate aerosol formation. Since both sulfate aerosol and CO₂ are known to regulate the radiation budget and solar energy balance on the Earth (Cheung et al., 2011; Möller, 1964), coupled with the CO₂-initiated promoted sulfate pathway found here, their overall influence on the global climate needs further investigation. Therefore, our study highlights the necessity for a comprehensive understanding of the CO${}_{\mathrm{3}}^{\mathrm{•}-}$-relevant chemistry in the underlying impacts of fine PM concentration, human health, and climate. All these assumptions need to be investigated in further detail. This study provides the indication that carbonate radicals not only play a role as a marginal intermediate in tropospheric anion chemistry but also as a strong oxidant for the surface processing of trace gas in the atmosphere.