As shown in Fig. S1 in the Supplement, OC was obtained with a custom combustion and sampling apparatus. The coal in the combustion experiment originated from Shanxi Province, China. The details of OC sampling and preparation are provided in Sect. S1.1 in the Supplement. The background concentrations of Fe, Mn and Cu in OC were measured through ICP-MS, and they varied in the range of 1.23–7.70 µg L⁻¹. The details of ICP-MS measurements are provided in Sect. S1.2. 1 mL of the extracted OC solution was dropped onto 6.0 × 1.5 cm prebaked quartz-fiber filters, and the solvents were dried under a N₂ stream (100 mL min⁻¹) in the dark at 298 K.

The sources and purities of salt chlorides used here are provided in Sect. S1.3. Stock solutions of metal salts (1.5 mM, in ultrapure water) were mixed with the OC solution at a 1:2 volume ratio, yielding mixtures containing 0.5 mM metal ions and 543.9 ± 35.4 µg C mL⁻¹ (Fig. S1). This metal-ion concentration was close to the levels reported in heavily polluted atmospheric conditions (Li et al., 2022). Each mixture was ultrasonicated for 30 min at 25 °C in the dark. The resulting products were designated as OC@TMIs, including OC@Fe³⁺, OC@Cu²⁺ and OC@Mn²⁺. Here, OC@TMIs referred to the actual sample prepared by introducing a specific TMI into OC. OC+TMIs was defined as a reference sample, where the interaction between OC and TMIs was assumed to be nonoccurrence. The property of OC+TMIs represented the arithmetic sum of that measured separately for OC and metal ion. Thus, OC+TMIs was utilized as a reference for evaluating synergistic or antagonistic effects of OC and TMIs. In the OC@NaCl control experiment, the NaCl concentration was 1.5 mM. 1.5 mL of the mixed solution was deposited on the inner surface of a quartz flow tube (20 cm length, 1.0 cm inner diameter). The flow tube was rotated slowly to ensure the uniform wetting of the surface. The coating was obtained by drying the samples under a N₂ stream (100 mL min⁻¹) in the dark environment at 298 K to prevent photochemical interference. Because a single sample could not be used sequentially for all experimental measurements, separate subsamples from the same initial batch of OC@TMIs were used to minimize potential variability.

The SO₂ uptake experiments were performed at ambient pressure in a horizontal cylindrical quartz flow tube reactor (34 cm length, 1.6 cm i.d.), as shown in Fig. S1. The details of the flow reactor and reaction procedure are given in Sect. S1.4. The changes in the SO₂ concentration were measured with a SO₂ analyzer (Thermo 43i). The calculation of the SO₂ uptake coefficient is described in Sect. S1.5.

Figure 1a–c illustrates the temporal evolution of SO₂ concentration on OC, TMIs and OC@TMIs under irradiation. All errors denoted the standard deviations of three independent replicates. Because chloride salts were used here, OC@NaCl control experiments were conducted. The decreasing trend of SO₂ concentration on OC@NaCl was the same as that on OC, suggesting no measurable effect of Cl⁻ on SO₂ uptake (Fig. S6). The SO₂ loss on FeCl₃ or MnCl₂ was less than that on OC, while it was similar for CuCl₂ and OC. When SO₂ was exposed to OC@Fe³⁺, the SO₂ concentration decreased by 1–2 ppb, which was obviously lower than that for OC+Fe³⁺. This indicates that the coexistence of Fe³⁺ with OC markedly suppressed the photochemical uptake of SO₂. The SO₂ loss on OC@Cu²⁺ and OC@Mn²⁺ was markedly greater than that on OC+Cu²⁺ and OC+Mn²⁺, respectively. This suggests that both Cu²⁺ and Mn²⁺ enhanced the photochemical reaction of SO₂ with OC. Dark control experiments showed that SO₂ uptake on OC@TMIs was slightly different from that on OC in the absence of light, indicating that the non-photochemical effects of TMIs to SO₂ uptake was limited (Fig. S7).

The initial (γi) and steady-state (γss) uptake coefficients of SO₂ are summarized in Fig. 1d. The γss was consistently lower than the γi due to the gradual aging and the depletion of photoactive components during the reaction. The γss on OC@Fe³⁺ only accounted for 5 % of that on OC+Fe³⁺, confirming that Fe³⁺ significantly restrained the photochemical SO₂ uptake on OC. In contrast, γss for OC@Cu²⁺ and OC@Mn²⁺ was 1.83 and 2.93 times of the corresponding one for OC+Cu²⁺ and OC+Mn²⁺, respectively. This definitely suggests that the coexistence of Cu²⁺ and Mn²⁺ with OC has synergistic effects on the photochemical uptake of SO₂, with Mn²⁺ exhibiting a stronger enhancement than Cu²⁺. Overall, these contrasting behaviors reveal that the addition of TMIs to OC can either restrict (Fe³⁺) or facilitate (Cu²⁺ and Mn²⁺) the photochemical uptake of SO₂, highlighting the distinct roles of TMIs in modulating the OC reactivity.

The contour maps of in-situ ATR-IR spectra of OC, TMIs and OC@TMIs exposed to SO₂ under irradiation are displayed in Fig. 2a. Several characteristic sulfate (SO42- or HSO4-) bands were identified for all samples in the 1300–1000 cm⁻¹ region (Zhang et al., 2023, 2019). It was noted that these bands intensified progressively with the time, reflecting the continuous accumulation of sulfates.

The IC results provided quantitative evidence of sulfate formation under irradiation (Fig. 2b). For OC@Fe³⁺, sulfate production was much lower than that on OC or FeCl₃. Specifically, the sulfate mass on OC@Fe³⁺ was merely 12 % of that for OC+Fe³⁺, confirming that Fe³⁺ restrained the formation of sulfates on OC. In contrast, the sulfate mass on OC@Cu²⁺ and OC@Mn²⁺ was 1.13 and 1.52 times larger than that for OC+Cu²⁺ and OC+Mn²⁺, respectively. These confirm that both Cu²⁺ and Mn²⁺ promote the photochemical oxidation of SO₂ to sulfates on OC. Only limited sulfate formation was observed on OC@TMIs under dark conditions, indicating a minor contribution from the non-photochemical pathways (Fig. S8). Taken together with the SO₂ uptake data in Fig. 1, TMIs exerted markedly different influences on SO₂ conversion to sulfates on OC.

The complexation between OC and TMIs may occur and alter the optical properties of OC (Wang et al., 2024). Changes in optical absorption and steady-state fluorescence signals were used to probe these interactions. Figure S9a–c displays the UV-vis spectra of OC, TMIs, OC+TMIs and OC@TMIs. OC and OC@TMIs exhibited a sharp increase in the absorbance at the shorter wavelengths. FeCl₃, CuCl₂ and MnCl₂ owned very weak absorbance. OC@Fe³⁺, OC@Cu²⁺ and OC@Mn²⁺ presented lower absorbance than OC+Fe³⁺, OC+Cu²⁺ and OC+Mn²⁺, respectively. This decrease in the absorbance suggests the occurrence of interactions between OC and TMIs. The background concentrations of Fe, Mn and Cu in OC were measured by ICP-MS, and they varied in the range of 1.23–7.70 µg L⁻¹. This accounted for only 0.0045 %–0.0276 % of TMI concentrations used for OC@TMIs. Therefore, the contribution of intrinsic Fe, Mn and Cu to the TMI dependent spectral changes was considered to be very limited. The FT-IR spectra of OC are displayed in Fig. S10, and the peak assignments are summarized in Table S1 in the Supplement. OC contained carboxyl, hydroxyl and carbonyl groups. These oxygen-containing groups with π bond electrons and lone pairs can serve as the complexation sites by donating electrons to the vacant orbitals of TMIs (Wang et al., 2021c). Coordination with TMIs can perturb the electronic structure of OC chromophores and affect π-π∗ or n-π∗ electronic transitions, which may induce a hypochromic effect in the UV-vis spectra (Qin et al., 2024). To further compare the spectral changes induced by TMIs, differential absorbance was calculated as ΔAbs = Abs(OC+TMIs) - Abs(OC@TMIs). Notably, as shown in Fig. S9d, there was the largest ΔAbs between OC@Fe³⁺ and OC+Fe³⁺ in the entire spectral range, indicating the strongest complexation of OC with Fe³⁺ among three TMIs.

EEM fluorescence spectroscopy combining with the PARAFAC analysis were employed to further investigate the complexation interactions between TMIs on OC. Three chromophores are identified in Figs. 3a–c and S11, and their peak positions and assignments are given in Table S2. C1 (Ex / Em = 265(320) / 380 nm) and C2 (Ex / Em = 285(370) / 420 nm) were assigned to chromophores with low and high oxidation states, respectively. C3 (Ex / Em = 200(260) / 340 nm) belonged to protein-like chromophores (Li et al., 2022; Wang et al., 2021b). Fluorophores at longer emission wavelengths, such as C2, typically owned abundant polar groups, including hydroxyl and carboxyl, which acted as primary binding sites for TMIs (Pan et al., 2020; Lu et al., 2019). The relative contributions of different components to OC shifted upon the addition of TMIs (Fig. 3d), reflecting the varied complexation strengths of TMIs with OC (Ma et al., 2022; Yan and Korshin, 2014). For OC@Fe³⁺ in comparison with OC, the ratio of C1 obviously increased by 14 %, while it decreased by 9 % and 5 % for C2 and C3, respectively. By contrast, OC@Cu²⁺ and OC@Mn²⁺ showed smaller changes in the C1–C3 proportions. These mean that Fe³⁺ forms the most stable complexes with OC due to its high charge density and trivalent oxidation state, which would enhance the ligand attraction and enable higher coordination numbers (Liu et al., 2022). Cu²⁺ and Mn²⁺, with lower charge densities and less favorable coordination geometries, exerted weaker complexation effects with OC.

As a comprehensive measure of fluorescence intensity, the total fluorescence volume (TFV) was used to evaluate the overall impact of metal complexations (Fig. 3e). The addition of TMIs led to a reduction in TFV, confirming the fluorescence quenching occurrence. Fe³⁺ caused the most pronounced reduction of TFV (85.2 %), followed by Cu²⁺ (13.4 %) and Mn²⁺ (1.4 %). This strong quenching effect was attributed to the high binding affinity of Fe³⁺ with highly oxidative and protein-like fluorophores. The complexation between OC and TMIs can disrupt the π-π∗ conjugation and deactivate fluorescence centers, which contributed to a significant reduction in TFV (Kuramochi et al., 2018). In addition, the complexation of Fe³⁺ with OC could promote the intersystem crossing (ISC) from the singlet state (¹OC^∗) to the triplet state (³OC^∗) (Ruzi et al., 2017; Treacy and Rovis, 2024), enhancing nonradiative decay and further suppressing the fluorescence emission.

To directly demonstrate the complexation strength between TMIs with species in OC, DFT calculations were conducted using Phenol (PH), benzoic acid (BA) and a model substance (MS) (Fig. S12). PH and BA were chosen according to FT-IR analysis (Fig. S10), as they represented species with hydroxyl and carboxyl groups commonly present in OC. MS simultaneously contained these functional groups with a π-conjugated aromatic backbone to better reflect the structural features of ambient OC. Section S3.1 and Fig. S13 illustrate the complexation structure of three molecular models with Fe³⁺, Cu²⁺ and Mn²⁺. As summarized in Table S3, all molecular models exhibited the strongest complexation energy with Fe³⁺, followed by Cu²⁺ and Mn²⁺. This was well consistent with the results observed in the UV-vis and fluorescence spectra. Although these models cannot capture the full molecular heterogeneity of OC, the consistency between the calculated results and experimental trends helped to understand the complexation effects of OC with TMIs. Complexation strength is commonly described using stability constants (log K). Because OC was a heterogeneous mixture with multiple potential binding sites, log K may not be exactly determined. Weak and strong complexation described here was relative sense, according to UV-vis spectral change, fluorescence quenching and complexation energy.

As described through Reactions (R1)–(R6), previous studies have proposed the photochemical conversion pathways of SO₂ to sulfates on OC (Yang et al., 2025, 2024a). Upon accepting photons, OC is excited to ³OC^∗, which subsequently generates free electrons (Reactions R1–R2) (Yang et al., 2024a; Wang et al., 2020). These electrons would reduce adsorbed O₂ to ⚫O2- (Reaction R3) (Zhang et al., 2022). ⚫O2- can combine with protons to form hydroperoxyl radicals (HO2⚫), which are converted into O₂ and hydrogen peroxide (H₂O₂) (Reaction R4) (Zhang et al., 2020c). H₂O₂ undergoes the photolysis to produce ⚫OH (Reaction R5), and ⚫OH drives the oxidation of adsorbed SO₂ or H₂SO₃ or HSO3- to sulfates (Reaction R6) (Liu et al., 2020b). R1OC→hv3OC∗R23OC∗→OC++e-R3O2+e-→⚫O2-R4⚫O2-+H+→HO2⚫→O2+H2O2R5H2O2→hv⚫OHR6⚫OH+SO2orH2SO3(HSO3-)→SO42-+H2O

To verify whether the above mechanism remained operative after the complexation between TMIs and OC, electrochemical and EPR analyses were conducted. Mediated electrochemical oxidation (MEO) data of OC without or with TMIs showed persistent occurrence of the photo-generated electron transfer (Figs. S14 and S15). BMPO was used as the spin-trapping agent. As displayed in Fig. 4a, the characteristic EPR signal of BMPO-⚫OH adducts was detected under irradiation, confirming the generation of ⚫OH. Moreover, EPR signals of BMPO-⚫OH adducts followed the order: OC@Mn²⁺ > OC@Cu²⁺ > OC > OC@Fe³⁺, which well aligned with γss and sulfate production. The critical role of ⚫OH in the conversion of SO₂ to sulfates was further determined using NaHCO₃ to scavenge ⚫OH (Yang et al., 2024a), where the sulfate formation was markedly restricted (Fig. S17). These demonstrate that ⚫OH generation capacity is the primary determinant of sulfate formation on OC@TMIs.

According to Reactions (R1)–(R2), the photogeneration of electrons was an initial step that can affect the reaction activity of OC. Figure 4b summarizes the distinct regulatory roles of Fe³⁺, Cu²⁺ and Mn²⁺ in the EDC of OC under irradiation. The EDC of OC@Fe³⁺ (0.00496 µmol e⁻ (mg C)⁻¹) was substantially less than that of OC+Fe³⁺ (0.01904 µmol e⁻ (mg C)⁻¹), suggesting that Fe³⁺ significantly suppressed the photogeneration of electrons. This mainly stemmed from the strong complexation of Fe³⁺ and OC, which masked the redox-active moieties and dynamically quenched ³OC^∗, restraining free electron formation (Li et al., 2022; Liu et al., 2024). Figure 4b shows that EDC of OC@Cu²⁺ (0.02074 µmol e⁻ (mg C)⁻¹) and OC@Mn²⁺ (0.02147 µmol e⁻ (mg C)⁻¹) was higher than that of OC+Cu²⁺ (0.01959 µmol e⁻ (mg C)⁻¹) and OC+Mn²⁺ (0.01812 µmol e⁻ (mg C)⁻¹), respectively. This means that Cu²⁺ and Mn²⁺ enhance the electron-donating ability of OC. These larger EDC may be related to reversible redox cycling of Cu²⁺ / Cu⁺ and Mn²⁺ / Mn³⁺ in the photochemical process. In contrast, Fe³⁺ formed more stable complexes with OC that trapped the photogenerated electrons, thereby suppressing efficient redox cycling of Fe³⁺ / Fe²⁺. Under irradiation, Cu²⁺ was reduced to Cu⁺ by accepting photogenerated electrons, which was subsequently re-oxidized to Cu²⁺ by O₂ or electron-deficient functional groups (e.g., quinones) (Pan et al., 2020). As displayed in XPS spectra of Cu 2p for OC@Cu²⁺ in Fig. 4c and d, the peaks at 932.5 and 934.3 eV corresponded to Cu⁺ and Cu²⁺, respectively (Biesinger et al., 2011). The proportion of Cu²⁺, denoted as R(Cu²⁺), decreased from 0.67 to 0.59 after photochemical SO₂ aging. This confirms the existence of reversible redox cycling of Cu²⁺ / Cu⁺. Similarly, Mn²⁺ can be oxidized to Mn³⁺ by photogenerated holes (OC⁺) or reactive oxygen-containing intermediates on OC, and then Mn³⁺ was reduced by phenolic or aromatic groups on OC (Hansard et al., 2011). The Mn 2p_3∕2 spectra of OC@Mn²⁺ exhibited two peaks at 640.4 and 641.9 eV (Fig. 4e and f), which were ascribed to Mn²⁺ and Mn³⁺, respectively (Cerrato et al., 2010; Xing et al., 2025). The proportion of Mn²⁺ (R(Mn²⁺)) decreased from 0.41 to 0.16 after the reaction, indicating Mn²⁺ / Mn³⁺ redox interconversion. These reversible redox cycles facilitated dynamic interfacial electron transfer by accepting and releasing photogenerated electrons. This bidirectional electron flow can amplify the outward electron flux from OC, thereby increasing the electron-donating capacity (Fulda et al., 2013; Li et al., 2021).

As shown in Reaction (R3), ⚫O2- formation relied not only on the photogenerated electrons but also on the O₂ activation. To elucidate the effect of TMIs complexation on the O₂ activation, DFT calculations were carried out in the interaction of O₂ with MS (Fig. S18). Figure 5a displays that O₂ adsorption energy on MS was -82.03 kJ mol⁻¹, with O–O bond length of 1.20 Å. For MS@Fe³⁺, O₂ adsorption energy became more negative (-205.59 kJ mol⁻¹), reflecting stronger electrostatic binding due to high positive charge of metal center (Zhang et al., 2024). However, O–O bond length remained unchanged, suggesting negligible electron transfer into the π∗ antibonding orbital of O₂. Thus, O₂ is strongly bound yet not appreciably activated. Complexation between OC with Cu²⁺ or Mn²⁺ obviously enhanced the adsorption of O₂ (-212.93 kJ mol⁻¹ for Cu²⁺ and -366.32 kJ mol⁻¹ for Mn²⁺). The O–O bond length slightly increased to 1.21 Å for MS@Cu²⁺, while it was markedly elongated to 1.32 Å for MS@Mn²⁺. These changes suggest electron transfer from the metal centers to the π∗ antibonding orbital of O₂, boosting effective activation of O₂ and facilitating the ⚫O2- generation. Compared to Cu²⁺, Mn²⁺ exhibited stronger O₂ activation, which was attributed to its half-filled 3d⁵ configuration and moderate redox potential, both of which favored efficient π∗ electron donation.

Accordingly, Fe³⁺ inhibited electron generation and failed to activate O₂, whereas Cu²⁺ and Mn²⁺ promoted these two processes, with Mn²⁺ showing a stronger enhancement effect. This led to the ⚫O2- signal intensity order of OC@Mn²⁺ > OC@Cu²⁺ > OC > OC@Fe³⁺ (Fig. 5b). ⚫O2- directly contributed to H₂O₂ formation (Reaction R4), which was quantified with a TiOSO₄ chromogenic method (Fig. S20). Figure 5c shows that H₂O₂ production trend had the consistency with ⚫OH generation ability among these samples (Fig. 4a), which ultimately governed the photoconversion of SO₂ to sulfates.

The OC samples were collected from smoke aerosols generated during coal combustion. The TMIs concentrations selected here were atmospherically relevant soluble metal ion levels. These systems may roughly represent the coexistence of OC and TMIs for atmospheric aerosols in polluted regions affected by coal combustion emissions. High humidity and acidic conditions may promote metal dissolution and facilitate the complexation between soluble metal ions and OC. The photochemical lifetime of SO₂ (τSO2) on OC@TMIs was calculated with Eq. (2) (Yang et al., 2021), 2τSO2=4γνA where γ represents the uptake coefficient of SO₂; ν denotes the mean SO₂ molecular velocity; A corresponds to the aerosol surface area concentration. Based on field measurements during heavily polluted episodes, A ranged from 5.44 × 10⁻⁵ to 8.16 × 10⁻⁵ cm² cm⁻³ (Zhang et al., 2018; Yang et al., 2021). Under typical atmospheric conditions (40 ppb SO₂, 60 % RH), γss for SO₂ on OC@TMIs were determined to be (0.39–22.24) × 10⁻⁶. Organic matters typically account for approximately 20 %–45 % of PM_2.5 mass (Wu et al., 2018). The τSO2 was estimated to be 1.8–379 d in the coexistence of OC and TMIs. By contrast, τSO2 on OC from coal combustion was much narrower, which ranged from 17 to 53 d under the same conditions (Yang et al., 2025).

To assess the atmospheric production of sulfates during the photooxidation of SO₂ on OC@TMIs, the sulfate formation rate (R) was estimated using Eq. (3), 3R=dSO42-dt=RPDg+4γν-1ASO2 where Dg refers to the diffusion coefficient of SO₂ (1.12 × 10⁻⁵ m² s⁻¹); RP denotes the aerosol particle radius and is estimated via Eq. (4) (Li et al., 2020), 4RP=0.254×PM2.5µgm-3+10.259×10-9 where [PM_2.5] denotes the mean mass concentration of PM_2.5, and a concentration range of 100–500 µg m⁻³ for the polluted episodes in typical Chinese cities was adopted here (Chu et al., 2020). Under these conditions, R was determined to be 0.07–5.95 µg m⁻³ h⁻¹, which was a substantially broader range than 0.43–1.33 µg m⁻³ h⁻¹ for OC (Yang et al., 2025). Broader ranges of τSO2 and R on OC@TMIs indicate that the coexistence of OC and TMIs significantly modulates the photochemical conversion of SO₂ to sulfates. These findings suggest that current model simulations may underestimate or overestimate sulfate formation on OC by overlooking such antagonistic or synergistic processes.

It has been well demonstrated that ⚫OH was the dominant active species responsible for the oxidation of SO₂ to sulfates on OC@TMIs (Figs. 4a and S17). Other metal ions (Ca²⁺, Co²⁺, Zn²⁺, Mg²⁺, Ni²⁺, Cr³⁺ and Al³⁺) commonly presenting in atmospheric aerosols were expanded to examine whether this mechanism exhibited broader generality (Hua et al., 2024). As shown in Fig. 6a, the peaks of BMPO-⚫OH adducts were observed under irradiation, confirming the ⚫OH generation on all OC@TMIs. Relative to OC, the coexistence of Ca²⁺, Co²⁺, Zn²⁺ and Mg²⁺ obviously promoted the ⚫OH generation, while Ni²⁺, Cr³⁺ and Al³⁺ exerted an inhibitory effect. Figure S21 summarizes the mass of sulfates formed on OC@TMIs, which was consistent with the ⚫OH generation trend. Importantly, a linearly positive relationship (R2 = 0.97) between ⚫OH intensity and sulfate mass (Msulfates) was determined in Fig. 6b. Msulfates can be parametrized as a function of ⚫OH signal intensity (I) with an equation: Msulfates = (1.68 ± 0.11) × 10⁻² × I - (0.43 ± 0.08) × 10⁻². The linear relationship between ⚫OH intensity and sulfate formation was established using fresh OC from coal combustion and different TMIs at the same concentration. OC in aerosols generally stems from diverse sources and undergoes the atmospheric aging processes, which would significantly modify the compositions and structures of OC (Cao et al., 2021; Deng et al., 2022). TMIs types and concentrations also display temporal and spatial variations (Ito et al., 2016; Moreno et al., 2011). These may alter the complexation between OC and TMIs and consequently affect ⚫OH generation and sulfate formation. Thus, there may be some uncertainties for the applicability of this linear relationship under ambient conditions. Nevertheless, this quantitative relationship highlighted the potential dual roles of metal ion types in the sulfate formation, challenging the traditional view that metal ions often promote the SO₂ oxidation (Cao et al., 2024a; Wang et al., 2022, 2021a). It also emphasized the critical role of ⚫OH in the sulfate formation on OC@TMIs. Accordingly, incorporating the ⚫OH generation ability of diverse OC@TMIs complexes into the atmospheric models may help to constrain sulfate source strength, predict haze evolution and assess associated health impacts in polluted environments.