Overlooked significance of iodic acid in new particle formation in the continental atmosphere

Significance

Iodic acid is a well-known driver for marine new particle formation (NPF), showing high efficiency in nucleation. Despite its recent detection over land and rapid global increase, its role in continental NPF and the resulting current and future atmospheric impacts remain unknown. Here, we present molecular-level evidence that iodic acid can substantially facilitate particle formation (1.5 to 50 times by 2060) by its synergistic nucleation with the typical land-based pollutants, sulfuric acid and amine, thus effectively capturing the observed inland NPF events in China. These findings highlight the significance of considering iodine-involved nucleation in accurately modeling continental particle formation and the resulting aerosol-related climatic impacts, particularly in the future with declining sulfur and increasing iodine emissions.

Abstract

New particle formation (NPF) substantially affects the global radiation balance and climate. Iodic acid (IA) is a key marine NPF driver that recently has also been detected inland. However, its impact on continental particle nucleation remains unclear. Here, we provide molecular-level evidence that IA greatly facilitates clustering of two typical land-based nucleating precursors: dimethylamine (DMA) and sulfuric acid (SA), thereby enhancing particle nucleation. Incorporating this mechanism into an atmospheric chemical transport model, we show that IA-induced enhancement could realize an increase of over 20% in the SA–DMA nucleation rate in iodine-rich regions of China. With declining anthropogenic pollution driven by carbon neutrality and clean air policies in China, IA could enhance nucleation rates by 1.5 to 50 times by 2060. Our results demonstrate the overlooked key role of IA in continental NPF nucleation and highlight the necessity for considering synergistic SA-IA-DMA nucleation in atmospheric modeling for correct representation of the climatic impacts of aerosols.

Atmospheric aerosols have substantial impacts globally on air quality, climate change, and human health (1–4). New particle formation (NPF), initiated by molecular clustering, is the major source of aerosols on a numeric basis (5–8). However, identifying molecular-level mechanisms in the key nucleation stage is challenging owing to the chemical complexity of the nucleation precursors and the influence of ambient conditions (6). Confirmed nucleation mechanisms primarily include sulfuric acid (SA)-driven nucleation with enhancement from basic gases (e.g., NH₃ and amines) (9–12), highly oxygenated organic molecules, and other precursors, in addition to the nucleation driven by highly oxygenated organic molecules themselves (13). Recent studies outlined a crucial role of iodine nucleation in driving marine NPF, with iodic acid (HIO₃; IA) identified as the predominant precursor among iodine species, e.g., iodine oxoacids (HIO_2-3) and iodine oxides (I₂O_3-5) (14–17). Evidence from CLOUD chamber experiments showed that IA can even form particles more efficiently than SA (18). Given that iodine emissions have tripled since the mid-20th century and are still rising (19, 20), iodine-associated impacts on NPF are expected to become further amplified over time.

In addition to the nucleation of homologous iodine species (14, 18, 21–23), pioneering studies proposed that IA can form stable clusters with typical NPF precursors such as SA (24, 25), NH₃ (26), and dimethylamine [(CH₃)₂NH; DMA] (27) individually, derived mainly from anthropogenic emissions. More importantly, IA is not only widespread over oceans but also detected at various continental sites (e.g., Beijing and Nanjing in China) with notable atmospheric concentrations (10⁶ to 10⁷ molec. cm⁻³) close to or even exceeding those of SA (18, 28), likely arising from the sea–land transport (28) and the recently unveiled land-based iodine emissions (29). Thus, given the combination of the existence of IA over land and its preponderant nucleation capacity, IA is highly likely to influence NPF in polluted continental atmospheres. Despite the prevalence of SA–DMA nucleation (9–11, 30), this binary mechanism falls short in explaining all the observed NPF rates (11, 31), indicating the presence of other as-yet unrevealed molecules in the process. Taken together, IA has great potential for participation in the SA–DMA nucleation process, thereby further influencing continental NPF. However, the underlying IA-involved multicomponent (SA–IA–DMA) nucleation mechanism and the resulting atmospheric impacts remain unexplored.

Here, we investigate the role of IA in the process of SA–DMA nucleation from microscale mechanisms to macroscale impacts by combining quantum chemical calculations, atmospheric cluster dynamic simulations, and Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) simulations. The results show that IA can structurally stabilize SA–DMA clusters via hydrogen bonds (HBs), halogen bonds (XBs), and protonating DMA, thereby kinetically facilitating nucleation. To probe the impacts of IA on atmospheric SA–DMA nucleation, we extended the WRF-Chem model to include a state-of-the-art iodine source, chemistry scheme, and SA–IA–DMA binary and ternary nucleation mechanism that effectively captures not only the field observations of IA distributions but also the NPF events. Our findings highlight the hitherto neglected importance of IA in continental NPF, thereby providing molecular insights into our understanding of atmospheric NPF over land regions.

Results

Molecular Evidence of IA-Involved SA–DMA Nucleation.

Atmospheric clusters form mainly via intermolecular interactions, e.g., HBs and XBs. To reveal the nature of SA–IA–DMA clustering at the molecular level, we probed their potential active bind sites and intermolecular interactions within the formed clusters by quantum chemical calculations. As shown in SI Appendix, Fig. S1, SA, IA, and DMA all possess HB donor and acceptor sites, with IA having additional XB sites, facilitating the formation of SA–IA–DMA clusters via both HBs and XBs. Nevertheless, precursor molecules can form complex cluster isomers with varying stability. We then adopted the multistep conformation search scheme (SI Appendix, section 1) to identify the low-lying structures for the (SA)_x(IA)_y(DMA)_z (1 ≤ z ≤ 3, 3 ≤ x + y + z ≤ 6) clusters. As shown in SI Appendix, Fig. S2, the pure acid clusters, i.e., (IA)_1-6 and SA–IA, are structurally stabilized by intermolecular O-H⋯O HBs and O-I⋯O XBs. The acid-base clusters, i.e., SA–DMA, IA–DMA, and SA–IA–DMA, additionally undergo proton transfer to yield cohesive ion pairs (SI Appendix, Fig. S3). Thermodynamically, the formation of acid-base clusters is favored more than that of pure acid clusters owing to the lower energy barriers (SI Appendix, Fig. S4). Within the SA–IA–DMA clusters, IA acts as a candidate for SA as a proton donor for DMA in forming DMAH⁺ and IO₃⁻. Notably, compared to the SA–DMA clusters, IA introduces more HBs (95.6% with medium strength, SI Appendix, Table S1 and section 2.3) and extra XBs within the SA–IA–DMA clusters, all of which can facilitate clustering. Consequently, SA–IA–DMA clustering can occur with a minor energy barrier of 1.24 kcal mol⁻¹, indicating its great potential for nucleation.

Nucleation Mechanism of the SA–IA–DMA System.

To explore how clusters kinetically form, as shown in Fig. 1, we tracked the detailed nucleation process of SA–IA–DMA system using the Atmospheric Cluster Dynamics Code (ACDC) at T = 280 K, CS = 1.7 × 10⁻² s⁻¹, [SA] = 3.4 × 10⁶ molec. cm⁻³, [IA] = 10⁵ to 10⁸ molec. cm⁻³, and [DMA] = 4.0 ppt, in accord with the observed concentrations (9, 14, 18, 28, 30). The formation of the nucleated cluster can proceed via SA–DMA, SA–IA–DMA, and IA–DMA pathways. The SA–DMA clusters are formed by sequential addition of SA and DMA monomers coupled with collision with (SA)₁(DMA)₁ heterodimers, consistent with the reported SA–DMA nucleation pattern (31, 32). The binary IA–DMA pathway mainly proceeds by the collision of IA or DMA monomer, forming the larger (IA)₄(DMA)₂ cluster, which can undergo further growth. In the SA–IA–DMA ternary pathway, the IA-involved formation of (SA)₁(IA)₁(DMA)_1-2 clusters is fundamental for yielding the larger (SA)₁(IA)₂(DMA)₃ and (SA)₂(IA)₁(DMA)₃ clusters with potential for further growth out of the simulation system (SI Appendix, Table S2 and section 3). The branch ratio of each clustering pathway varies with precursor concentrations. At lower [IA] (≤10⁶ molec. cm⁻³) than [SA] (3.4 × 10⁶ molec. cm⁻³), SA–DMA clustering is dominant (67 to 94%). As [IA] increases to 10⁷ molec. cm⁻³, while holding other parameters constant (i.e., T, CS, [SA], and [DMA]), SA–IA–DMA synergistic clustering dominates nucleation (85%, Fig. 1B). Under the conditions of higher [IA] (10⁸ molec. cm⁻³), the IA–DMA pathway contributes most (69%). When IA is less abundant, i.e., comparable to or lower than [SA], it acts as an enhancer of SA–DMA nucleation; conversely, it can even supersede SA as a clustering driver. Thus, predictably, IA enhancement on nucleation is substantial, especially in regions with notable levels of IA and DMA, scarce SA, and lower temperatures (SI Appendix, Figs. S5–S7). Furthermore, as shown in Fig. 1C, we also examined the resulting cluster formation rate (J) as well as the IA’s enhancement on J (R = (J_SA-IA-DMA − J_SA-DMA)/J_SA-DMA × 100%) at varying concentrations. As [IA] increases from 10⁵ to 10⁸ molec. cm⁻³, R accordingly rises from 4% to over 24,000%, indicating that IA has a substantial promotional impact on SA–DMA nucleation. Even at relatively low [IA] (10⁶ molec. cm⁻³), IA can boost J by ~52% at 280 K, at which point the direct contribution of the IA-involved pathway to nucleation is 26% (Fig. 1B). Moreover, IA can also indirectly “catalyze” the SA–DMA cluster formation by first participating in yielding (SA)₂(IA)₁(DMA)₂ and then evaporating out to yield (SA)₂(DMA)₂ cluster. Consequently, the dual role of IA enhances the overall J.

Fig. 1.

Nucleation mechanism of the SA-IA-DMA system. (A) Clustering pathways, (B) contribution of pathway to flux out, and (C) enhancement of cluster formation rate (J, cm⁻³ s⁻¹). Solid and dashed arrows in (A) link the growth step of clusters via collision with monomers or clusters, respectively. And the forked arrow indicates the evaporation of monomers from the cluster. The bar graph in (C) indicates the enhancement of IA on nucleation (R = (J_SA-IA-DMA − J_SA-DMA)/J_SA-DMA × 100%). Simulations in (B and C) were conducted at T = 280 K, CS = 1.7 × 10⁻² s⁻¹, [IA] = 10⁵–10⁸, [SA] = 3.4 × 10⁶ molec. cm⁻³, and [DMA] = 4.0 ppt. Numbers in (B) indicate the contribution of the clustering pathways.

Validation of Three-Dimensional Model Simulations through Comparison with Observations.

The revealed nucleation mechanism of SA–IA–DMA system here was further implemented in the WRF-Chem model to probe its impact under ambient conditions. Primarily, IA was included in this WRF-Chem model based on comprehensive representation of atmospheric iodine sources, sinks, and chemistry, including the latest insights into the IA formation mechanism (29, 33). Then, the SA–IA–DMA binary and ternary nucleation was parameterized and introduced in the WRF-Chem model using a lookup table derived from over five million ACDC simulations (refer to Materials and Methods). Here, East Asia was selected as the target region, and China was considered for particular focus owing to the observed intensive NPF events with high nucleation rates (9, 10, 34). The simulated concentrations of nucleating precursors and particle number size distributions (PNSDs) were compared with the site observations to ensure the reliability of our updated model.

Simulated IA concentrations and observations at two inland sites in Beijing and Nanjing during both winter and summer are compared in Fig. 2A. Generally, the updated model incorporating iodine species reproduces the diurnal trend and peak daytime concentrations of IA at both sites and in both seasons. The mean IA concentrations for the simulated period between the simulations and observations are within a factor of 2 (SI Appendix, Table S6). Moreover, the observed features such as higher IA concentrations in summer than in winter at the both sites, higher IA concentrations in winter in Nanjing compared to those in Beijing, and the delayed timing of peak concentrations at the Beijing site during winter compared to that in summer are effectively captured by the model. Simulated nighttime IA concentrations are lower than the observed values in both summer and winter, despite the consideration of recently proposed active nighttime atmospheric iodine chemistry (35). Nighttime underestimation has also been observed in a recent box model simulation conducted by Finkenzeller et al. (33) based on the latest IA chemistry, which implies that currently unrevealed iodine chemistry might maintain iodine chemical cycling even under conditions of limited illumination. The underestimation of simulated IA concentrations during nighttime, together with the daytime bias compared to the observations, could also be attributed to existing uncertainties associated with multiple sources and sinks of iodine species, particularly in the continental atmosphere, which merit further investigation. Notably, however, the nighttime underestimation of IA concentrations did not affect the NPF simulations because the model generally captures the IA concentrations during daytime when NPF events typically occur. Additionally, the model effectively reproduces the concentrations of SA and DMA in Beijing within a factor of 1.7 (SI Appendix, Figs. S8 and S9 and Table S6). Given the difficulty in accurately simulating the concentrations of these nucleation precursors reported in previous related studies (36, 37), our model demonstrates reasonable performance.

Fig. 2.

Comparison of WRF-Chem simulations and observations. (A) IA concentration in Beijing (Beijing University of Chemical Technology; BUCT) and Nanjing (Station for Observing Regional Processes of the Earth System; SORPRES) during winter and summer. Winter months encompass December 2019, January 2020, and February 2020, while summer months comprise June, July, and August 2019. Observed IA concentrations are adopted from Zhang et al. (28). Simulated IA concentrations are averaged over nine model grids surrounding the observational sites. (B) Particle number size distributions (PNSDs) at Beijing (BUCT) in January and August 2019; the Left side is the averaged PNSDs and the Right side is the time series of the PNSDs. “Obs” refers to the field observations, and “Sim_MechNoSA-IA-DMA” indicates WRF-Chem simulations without the involvement of the SA-IA-DMA binary and ternary nucleation.

The simulated and observed PNSDs are further compared in Fig. 2B. The updated model, incorporating SA–IA–DMA binary and ternary nucleation and other nucleation mechanisms already present in the model (Materials and Methods), effectively captures the NPF events, the nucleation rate, and the averaged PNSDs at the Beijing site during the simulated period. On the contrary, simulations excluding SA-IA-DMA binary and ternary nucleation exhibit a significant deviation from observational data (SI Appendix, Figs. S10 and S11). With reasonable representation of precursor concentrations, this highlights the reliability of our parameterization scheme in calculating SA–IA–DMA binary and ternary nucleation rates based on molecular-level kinetic simulations using ACDC. The disparity in the averaged PNSDs between the simulated and observed data in the 2 to 10 nm range primarily originates from simplified assumptions made with respect to particle growth simulations, as discussed in our previous study (38).

Contribution of IA to NPF over China.

We selected January and August 2019 as representative months for winter and summer, respectively, to further investigate the impact of IA on NPF under different atmospheric conditions. The spatial distribution of near-surface IA concentration in January and August is presented in Fig. 3A. Overall, the concentration of IA exhibits a trend of being higher over the sea than over land, and higher during summer compared to winter, consistent with observed features (28). The subsided continental IA concentration in comparison to the maritime context should originate from the elevated particulate matter and NO_x (primarily NO₂) levels over land in diminishing the formation of IA. Specifically, elevated levels of NO₂ divert iodine chemistry into preferential formation of iodine nitrates rather than into the favored formation of iodine oxides, the precursors of IA, at low NO₂ concentrations (SI Appendix, Table S4) (39). Moreover, high concentrations of particulate matter over land deplete IA precursors by aerosol uptake, lowering IA concentration. In mainland China, while IA concentrations are notably lower than those over the ocean, eastern and southern coastal regions and some southwestern inland regions exhibit notably higher IA concentrations than other regions during both winter and summer, reaching levels up to 10⁶ and even 10⁷ molec. cm⁻³. This is attributed to more active oceanic transport and relatively low concentrations of PM_2.5 and NO₂ in these specific regions. At such levels of concentration, close to that of SA (SI Appendix, Fig. S8), IA could play an active role in SA–DMA nucleation by enhancing its nucleation rate.

Fig. 3.

Spatial distribution of IA concentration and enhancement of IA on SA–DMA-based nucleation in January and August 2019. (A) Spatial distribution of IA concentration, (B) enhancing ratio (ER) [(scenario with IA − scenario without IA)/scenario without IA)] with current anthropogenic emission, and (C) ER with future anthropogenic emissions under carbon neutrality and clean air policies (emissions of SO₂, NO_x, PM_2.5, VOCs, and NH₃ reduced by 96%, 96%, 96%, 80%, and 32% from current levels, respectively). Note that the color bar ranges in (B) and (C) are different.

As shown in Fig. 2 and SI Appendix, Figs. S10 and S11, the SA-IA-DMA binary and ternary nucleation exerts significant nucleation rates over China, with eastern regions identified as NPF hotspots. The contributions from IA to original SA–DMA nucleation can be assessed through the enhancing ratio (ER) based on a set of comparative NPF scenarios with and without IA involvement. The ER is defined as follows: (scenario with IA − scenario without IA)/scenario without IA × 100%. The process via which IA enhances the SA–DMA nucleation rate in January and August is illustrated in Fig. 3B. As expected, IA causes notable enhancement of the SA–DMA nucleation rates in most marine atmospheres with an ER of >100%, albeit with the absolute nucleation rates lower by several orders of magnitude than those over land (SI Appendix, Fig. S11). In mainland China, regions with a high ER align with regions with elevated IA concentrations (Fig. 3A), low SA concentrations (SI Appendix, Fig. S12), and considerable DMA levels (SI Appendix, Fig. S13). In winter months, the ER generally exceeds 5% in coastal regions, reaching over 10% in eastern regions. In summer months, the ER is similar to those in winter months in most continental regions, but it can exceed 20% in northeastern and southeastern regions and 1,000% in southwestern regions. Of note, while the IA levels are generally higher in summer compared to those during winter, the resulting IA’s enhancement on NPF in winter and summer is comparable in most regions of China. This arises from the suppressed participation of IA in SA–IA–DMA ternary nucleation with elevated summer temperatures and SA concentrations. The notable IA enhancements in high-elevation southwestern and high-latitude northeastern regions during summer are a result of high IA concentrations, low temperatures, and low SA concentrations. In summary, these results suggest that continental NPF is likely to involve the participation of IA in a widespread manner, and in some regions, IA might play a crucial role. However, the contribution from IA to continental NPF has long been overlooked.

Given the objectives of carbon neutrality and further air quality improvements, by 2060, it is highly likely that mainland China’s anthropogenic emissions of pollutants, including SO₂, NO_x, PM_2.5, and volatile organic compounds (VOCs), could be reduced substantially to synergistically mitigate climate change and air pollution. Thus, the condensation sink (CS) coefficient is expected to decline in the future (SI Appendix, Fig. S14). Reductions in future anthropogenic emissions and the consequential lowering of concentrations of SA, PM_2.5, and NO₂ are highly likely to magnify the enhancing effect of IA on SA–DMA nucleation. Therefore, we further probed IA enhancement under a future scenario where anthropogenic emissions of PM_2.5, VOCs, NO_x, SO₂, and NH₃ are reduced by 80%, 96%, 96%, 96%, and 32%, respectively. These emission reduction ratios for pollutants were derived from a comprehensive model framework that considers prospective low-carbon and pollution control measures by 2060 (see details in Materials and Methods). As shown in Fig. 3C, under future emissions, the enhancing effect of IA on SA–DMA nucleation is notably larger than that at present during both winter and summer. The ER is higher than 200% in winter and 50% in summer across most regions in China. In the hotspot regions of China where IA contributes to nucleation, the SA–DMA nucleation rates during winter months are projected to be enhanced by a factor of 20 in the southeastern and the southwestern regions; in the summer months, the SA–DMA nucleation rates are expected to be enhanced by a factor of 5 in the southeastern region and by a factor of >50 in the southwestern and northeastern regions. Note that here the future iodine emissions are set to be the same as the current levels. Given that existing field evidence suggests that future iodine emissions will continue to increase (19, 20, 40, 41), the future augmentation of the ER in this study might be underestimated. Therefore, under the driver of future pollution reduction efforts, the role of IA in continental NPF will rise rapidly, potentially dominating the nucleation processes.

Uncertainty Analysis of Simulations.

During simulations, the present IA’s impacts on the SA–DMA nucleation as obtained in this study might be subject to uncertainties from both macroscale and microscale simulations. As for the macroscopic aspect, the 3-D representation of the key factors determining IA enhancement on SA–DMA nucleation, such as precursor concentrations and PNSDs, is constrained through evaluation against observations. However, despite employing the most recommended theoretical approach, the quantum calculations introduce uncertainties in depicting the thermodynamics of SA–IA–DMA clusters. This uncertainty poses a primary challenge in validating the impact of IA on SA–DMA nucleation. Thus, to ensure the robustness of our results, we probed the impact of uncertainties arising from quantum chemical calculations on both ACDC and WRF-Chem simulations, as shown in SI Appendix, Figs. S15–S18, by examining the effect of altering the employed ΔG of all clusters (i.e., the final results of quantum chemical calculations) by ±1 kcal mol⁻¹ (reflecting the currently realized uncertainty range in ΔG) or ±2 kcal mol⁻¹ (an extreme case exploring potential uncertainty not yet realized, see SI Appendix, section 4) compared to the base scenario on the resulting IA enhancement on the SA–DMA nucleation rate. Consequently, the ACDC simulations in SI Appendix, Figs. S15 and S17 show that the uncertainty in IA enhancement has a margin of error of 1.3 to 11.9% in typical inland regions, e.g., Beijing, with relatively low [IA] (2.0 × 10⁶ molec. cm⁻³). Moreover, this uncertainty increases (4.9 to 21.7%) in regions with richer IA ([IA] = 5.0 × 10⁶ – 1.0 × 10⁷ molec. cm⁻³). Further WRF-Chem simulations under atmospheric conditions (SI Appendix, Figs. S16 and S18) suggest that IA enhancement in both summer and winter slightly decreases or increases, while generally retaining the same overall pattern. We thus speculate that uncertainty in the quantum chemistry calculation does not alter the findings regarding IA enhancement of SA–DMA nucleation. Additionally, we also examined other input parameters that could potentially introduce modeling uncertainties, e.g., CS and the collision enhancement factor (EF). After testing in SI Appendix, Figs. S19 and S20, their impact on IA enhancement was found relatively limited and does not alter our main findings.

Discussion

In this work, we provide evidence that IA can structurally stabilize the SA–DMA clusters, thereby kinetically facilitating nucleation. We further combined the ACDC and WRF-Chem models to reveal the enhancing effects of IA on SA–DMA nucleation under atmospheric conditions. Consequently, the “bottom–up” atmospheric modeling framework that we developed highlights the as-yet unrecognized notable role of IA in NPF with key land-based NPF precursors (i.e., SA and DMA), which necessitates comprehensive representation of the iodine-driven synergistic nucleation in modeling continental NPF, as conducted in this study, over both regional and global scales.

Previous studies reported some field measurements of IA (14, 16, 18, 28), but details of its regional distribution remain unclear or missing. Our WRF-Chem model generally captures the spatiotemporal distribution of IA, e.g., the diurnal and seasonal variability (28). Despite a slight underestimation of nighttime IA concentrations in Beijing and Nanjing (Fig. 2A), likely attributable to some unidentified iodine sources, IA concentrations during daytime (typical NPF period) are effectively captured. Accordingly, the IA-involved PNSDs simulated by the extended WRF-Chem model also effectively fit with the field observations. Such enhancement of IA on NPF shows distinct correlation with IA distribution, with greater impact in coastal areas of southeastern China, such as Zhejiang (SI Appendix, Fig. S21), aligning with the field-based findings reported by Yu et al. (15) that indicate the important contribution of IA in the nucleated particles. Our findings advance understanding of IA’s contribution in the inland NPF process, not only during the previously reported early growth (1.8 to 3 nm) stage (28) but also during the key nucleation stage (i.e., initial formation of particles below 1.8 nm). Recent research (25) showed that IA can enhance SA particle nucleation in marine or polar regions; our study further found that IA can enhance SA-DMA particle nucleation over land. Nevertheless, given the atmospheric complexity, joint nucleation involving other potential precursors such as nitric acid (42) and iodous acid (HIO₂) (18, 25) might also exert influence concurrently. Therefore, such potential multicomponent synergistic nucleation deserves further exploration in future studies.

These findings from the microscale mechanism of nucleation to the resulting atmospheric effects hold important atmospheric implications: i) providing insight into the molecular-level nucleation dynamics of IA and SA-DMA; ii) illustrating that the simulated distribution of IA by our model fills gaps in rarely available observation data, establishing a paradigm for predicting broader and even global distributions; and iii) identifying that IA-involved synergistic nucleation forms particles efficiently, not only by capturing recent inland NPF events but also by facilitating accurate prediction of historical and future aerosol-related climatic impacts. In the future, as concentrations of anthropogenic pollutants decrease and iodine levels rise, iodine-mediated NPF is projected to exert progressively more important impact on global climate forcing in the 21st century.

Materials and Methods

Cluster conformation calculations were performed to identify stable structures from possible cluster isomers. The wavefunction analysis reveals the clustering nature by analyzing intermolecular interactions, and ACDC simulations kinetically illustrate the cluster formation process.

Cluster Structure Calculations and Analysis.

To fully sample isomers of the (SA)_x(IA)_y(DMA)_z (1 ≤ z ≤ 3, 3 ≤ x + y + z ≤ 6) clusters, a multistep conformational search scheme in previous related studies (27, 43) was adopted here (SI Appendix, section 1). The (IA)_1-6, (IA)_m(DMA)_n, (SA)_m(DMA)_n (n ≤ m, m + n ≤ 6), and (SA)_p(IA)_q (2 ≤ p + q ≤ 6) cluster structures used here referred to other previous related studies (24, 27, 43, 44) and were reoptimized in this study. Here, acid-base clusters with more bases than acids were not considered owing to their instability (27, 32). All identified stable clusters were finally optimized with tight convergence using the Gaussian 09 package (45) at the ωB97X-D/6-311++G(3df,3pd) (for H, C, N, O, and S) + aug-cc-pVTZ-PP (for I) (46, 47) level of theory, owing to the success of ωB97X-D functional in yielding the credible molecular clusters (48). Cartesian coordinates of the resulting final structures are listed in SI Appendix, Table S7. Further single-point energy was calculated with tight SCF at the DLPNO-CCSD(T)/aug-cc-pVTZ-PP (TightPNO) level of theory by ORCA 5.0 program (49). Notably, despite the accuracy of the current method, a prior benchmark study indicated that DLPNO-CCSD(T) results exhibit potential errors of less than ~1 kcal mol⁻¹ compared to the more accurate but expensive CCSD(F12*)/CBS results (48). Herein, the Gibbs free energies (ΔG_ref) of cluster formation, calculated by Eq. 1, are listed in Dataset S1:

$${\mathrm{\Delta }G}_{\mathrm{r}\mathrm{e}\mathrm{f}}={\mathrm{\Delta }E}_{\mathrm{D}\mathrm{L}\mathrm{P}\mathrm{N}\mathrm{O}-\mathrm{C}\mathrm{C}\mathrm{S}\mathrm{D}\left(\mathrm{T}\right)}+{\mathrm{\Delta }G}_{\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l}}^{\omega \mathrm{B}97\mathrm{X}-\mathrm{D}},$$ [1]

where ΔG_thermal and ΔE_{DLPNO-CCSD(T)} are the thermal and electronic contributions to ΔG_ref, respectively, and ΔG_ref at varying temperatures (T = 200 to 320 K, interval: 5 K) was calculated using the Shermo 2.0 code (50). Given the actual vapor pressures of each studied precursor, ΔG_ref can be further converted to ΔG(P₁, P₂,…, P_n) using Eq. 2 (32):

$$\mathrm{\Delta }G\left(P_1,P_2,\cdots ,P_n\right)=\mathrm{\Delta }{G}_{\mathrm{ref}}-k_{\mathrm{B}}T{\displaystyle \sum _{i=1}^n}{N}_i\mathrm{In}\left(\frac{P_i}{P_{\mathrm{ref}}}\right),$$ [2]

where k_B denotes the Boltzmann constant, N_i indicates the number of component i within the cluster, and P_ref and P_i signify the reference pressure (1 atm) and the actual pressure of vapor i, respectively.

To comprehend the bonding nature within the identified clusters, the wavefunction analysis based on quantum chemical calculations was performed by Multiwfn 3.7 (51) (SI Appendix, Table S1). More computational details are provided in SI Appendix.

ACDC Simulations.

To explore nucleation kinetics, the cluster formation rates, the clustering pathways, and the steady-state concentrations were calculated through precisely solving the birth–death equations (Eq. 3) using the ACDC (52) without any fitted parameters:

$$\begin{array}{c}\frac{{\mathit{dC}}_i}{\mathit{dt}}=\frac{1}{2}{\displaystyle \sum _{j<i}}{\beta }_{j,\left(i-j\right)}{C}_j{C}_{\left(i-j\right)}+{\displaystyle \sum _j}{\gamma }_{j,\left(i-j\right)\to i}{C}_{\left(i+j\right)}\hfill \\ -{\displaystyle \sum _j}{\beta }_{i,j}{C}_i{C}_j-\frac{1}{2}{\displaystyle \sum _{j<i}}{\gamma }_{i\to j}{C}_i+Q_i-S_i,\hfill \end{array}$$ [3]

where subscript i refers to the index of monomer i or cluster i, β_i,j and γ_i,j are the rate coefficients of collision and evaporation, respectively, C_i indicates cluster concentration, and Q_i and S_i are the terms of source and sink, respectively. Further details of the calculation of β_i,j and γ_i,j are summarized in SI Appendix. And the performed ACDC simulations considered all possible processes of collision and evaporation in clustering. Moreover, an enhancement on β_i,j attributable to intermolecular van der Waals forces is considered here, and the employed factor of 2.3 fits with the results of atomistic simulation (53), which succeeded in other clustering studies (9, 38, 54). Additionally, the sink term S_i for each molecule/cluster adopted a size-dependent sink as follows:

$$Si={\mathrm{C}\mathrm{S}}_{\mathrm{r}\mathrm{e}\mathrm{f}}\times {\left(d_i/d_{\mathrm{r}\mathrm{e}\mathrm{f}}\right)}^{-m},$$ [4]

where d is the diameter of the cluster or molecule, CS_ref is the condensation sink (CS) of the reference monomer (i.e., the SA monomer in this study), and m, a power law exponent (1.7), depends on scavenger distribution, which falls within the typical range for atmospheric aerosols (55). The values of CS range from 5 × 10⁻⁵ to 5 × 10⁻¹ s⁻¹, thereby covering reasonably clean and polluted environments (56). Generally, environments with high CS suppress nucleation (SI Appendix, Fig. S22).

Incorporating SA–IA–DMA Binary and Ternary Nucleation into WRF-Chem.

Owing to the involvement of multiple molecules/clusters in collisions and evaporation within the SA–IA–DMA nucleation system, and the variation of nucleation pathways under different conditions, it is not practically feasible to represent the SA–IA–DMA binary and ternary nucleation rate in a 3-D model using explicit mathematical expressions. Here, we built a SA–IA–DMA nucleation rate lookup table based on more than five million (5.32 × 10⁶) ACDC simulations with varying precursor concentrations ([SA] = 10⁵–10⁹, [DMA] = 5 × 10⁵ to 5 × 10⁹, and [IA] = 10³–10⁹ molec. cm⁻³) and atmospheric conditions (T = 200 to 320 K, CS = 5 × 10⁻⁵ to 5 × 10⁻¹ s⁻¹). As presented in Dataset S2, the range of variation for each parameter generally covers typical atmospheric values. The SA–IA–DMA binary and ternary nucleation rate was calculated online in WRF-Chem using the lookup table through a multivariable interpolation scheme following ref. 57.

Even before adding SA–IA–DMA binary and ternary nucleation, seven nucleation mechanisms (three mechanisms for organic nucleation: ion-induced/neutral nucleation driven by pure-organics, and nucleation driven by SA and organics; four mechanisms for inorganic nucleation: binary ion-induced/neutral nucleation driven by SA and H₂O, and ternary ion-induced/neutral nucleation driven by NH₃, SA, and H₂O) were already integrated into the initial version of the WRF-Chem model (58). Organic-SA nucleation is driven by organic compounds (O:C > 0.4) with ultralow and extremely low volatility (ULVOCs and ELVOCs), originating from the oxidation of monoterpenes, while pure-organic neutral/ion-induced nucleation is driven by ULVOCs with O:C > 0.4. The model traces the volatility distribution and formation chemistry of monoterpenes using the R2D-VBS framework (58), with experimentally constrained parameters. Because R2D-VBS species can introduce notable computational burden, the R2D-VBS framework is implemented in WRF-Chem as an equivalent one-dimensional VBS by summing all species with different O:C but with the same room-temperature volatility. Further details regarding the R2D-VBS framework and its implementation in WRF-Chem can be found in our previous study (58). The ACDC-derived parameterization used in this study is broadly consistent with a previously employed parameterization based on simplified cluster dynamics under most atmospheric conditions. Furthermore, it predicts nucleation rates at high temperatures (>310 K) in a more reasonable manner.

Representation of Source/Sink of DMA in WRF-Chem.

We calculated the DMA concentrations by implementing sources and sinks in WRF-Chem, the details of which can be found in our previous study (38). Briefly, owing to lack of a bottom–up DMA emission inventory, DMA emission was estimated by combining NH₃ emissions and DMA/NH₃ ratios. The original source-dependent DMA/NH₃ emission ratios were derived from simultaneous observation of NH₃, C1–C3 amines, NO_x, and SO₂, together with meteorological conditions observed at a suburban location in Nanjing (59); for marine DMA emission, the NH₃ emission was adopted from ref. 60, and the DMA/NH₃ emission ratio was derived from a recent campaign conducted in offshore areas of China by Chen et al. (61). The depletion of DMA concentration by gas-phase oxidation, wet deposition, and aerosol uptake was explicitly represented in the model using the key parameters listed in SI Appendix, Table S3.

Representation of Source/Sink of HIO₃ in WRF-Chem.

We incorporated 15 iodine species in WRF-Chem, including inorganic I₂, HOI, HI, I, IO, OIO, IOIO, I₂O₃, I₂O₄, IOIO₄, HIO₃, INO, INO₂, IONO₂, and organic CH₃I. Both HOI and I₂, emitted in the marine environment, are regarded as major natural sources of atmospheric iodine species (62). In this study, marine emissions of HOI as well as I₂ were calculated online based on surface wind speed, surface O₃ concentration, and sea surface temperature, following the parameterization from ref. 63. Another natural iodine source species, i.e., CH₃I, which is emitted from rice paddies, biomass burning, wood fuel, and wetlands, was taken from ref. 64. We note that a recent study highlighted that the intensity of anthropogenic iodine emission can even exceed natural sources in certain regions such as China and India (29). Thus, we also included the anthropogenic emissions of HI and I₂ of mainland China based on the emissions factor method following ref. 29. Specifically, the anthropogenic iodine emission inventory includes HI and I₂ emission from coal combustion. A comprehensive activity dataset of coal consumption from power plants, industry, and residential burning was used (65). Emission factors of gaseous HI and I₂ were calculated by combining the measured iodine content in coal and the removal efficiencies of air pollution control devices (66). The proportions of emitted HI and I₂ were set at 95% and 5%, respectively, according to measurements from ref. 67.

After the primary emission, the remaining iodine species including HIO₃ are formed through iodine chemistry processes, including gas-phase radical reactions, thermal decomposition reactions, and photochemical reactions (SI Appendix, Table S4). The iodine chemistry included in the model is primarily based on a box modeling framework developed recently by Finkenzeller et al. (33), together with some additional reactions reported previously (SI Appendix, Table S4). In the updated model, HIO₃ is formed not only from the well-established reaction of OIO + OH but also from a missing mechanism of IOIO → IOIO₄ → HIO₃ proposed by Finkenzeller et al. (33). Additionally, iodine species can be removed from the atmosphere through wet deposition and uptake by preexisting aerosols. In addition to the consumption of iodine, recent studies have revealed that recycling of iodine species through heterogeneous reactions involving gaseous HOI, INO₂, and IONO₂ and particulate IO₃⁻ during aerosol uptake can substantially regulate iodine species partitioning in gas and particle phases, despite the mechanism being uncertain (68–70). Here, we considered the conversion of three reservoir species (i.e., HOI, INO₂, and IONO₂) into reduced iodine species IX (X = Br/Cl) after aerosol uptake following ref. 71. Because other halogens are not involved in the model, we assumed IX as I₂ given the similar abilities of these species in the iodine cycle (71). For particulate IO₃⁻, we assumed that after condensation, HIO₃ is instantaneously reemitted to the atmosphere as IO following ref. 33. The key parameters for iodine species depletion including Henry’s law constants and uptake coefficients are summarized in SI Appendix, Table S5.

WRF-Chem Configuration.

We incorporated the SA–IA–DMA binary and ternary nucleation mechanism to the WRF-Chem model to investigate the influence of IA on atmospheric NPF. A simulation domain covering eastern Asia (horizontal resolution: 27 km), which has considerable IA concentrations and intense NPF events, was chosen here. For vertical resolution, there are 24 layers from surface to 50 hPa, including more layers at lower heights. The ABaCAS-EI 2017 and IIASA 2015 emission inventories were applied to China and other regions, respectively (72, 73). The MEGAN v2.04 model was employed to calculate biogenic emissions (74). For representing the future emission scenario of 2060, we cut all anthropogenic emissions of SO₂, NO_x, PM_2.5, VOCs, and NH₃ by 96%, 96%, 96%, 80%, and 32% from the current levels, respectively. The reduction ratios were estimated based on a comprehensive framework designed to achieve carbon neutrality and air quality improvement through available low-carbon and air pollution control measures. The changes in anthropogenic emissions between current levels and 2060 closely resemble those estimated in several previous studies that considered carbon neutrality and clean air policies in China (75–77).

Apart from tracking the gas and aerosol chemistry associated with monoterpenes via the R2D-VBS framework, the simulation relied on the SAPRC-99 gas chemistry scheme in tandem with the MOSAIC model and a one-dimensional VBS to simulate all other gas and aerosol chemical processes (78). National Centers for Environmental Prediction Final Analysis reanalysis data were used for the meteorological initial and boundary conditions, with spatial and temporal resolution of 1.0° × 1.0° and 6 h, respectively. The simulation results of the CAM-Chem model were employed as the chemical initial and boundary conditions. The primary physical options used in the simulations included the Morrison 2-moment cloud microphysics scheme, RRTMG shortwave and longwave radiative transfer schemes, the Eta similarity surface-layer scheme, unified Noah land-surface model, Mellor–Yamada–Janjic (Eta) TKE scheme, and Grell–Freitas ensemble scheme. More details about the emission inventory and the model used are provided in SI Appendix.

Our simulation comprised three parts: the simulation period for IA concentration that included December 2019, January 2020, and February 2020 (winter) and June, July, and August 2019 (summer). The periods were set to match the time periods of IA concentration observations conducted by Zhang et al. (28) in Beijing and Nanjing. For simulation of the impact of IA on NPF under current anthropogenic emissions, we selected January and August 2019 to represent winter and summer, respectively. These time periods also correspond to the available particle size distribution data at the Beijing site. To investigate the performance of the model simulation before and after incorporating our SA-IA-DMA system, we set up two parallel simulations: one considering both the SA-IA-DMA binary and ternary nucleation and other mechanisms (Sim_MechAll), and the other considering only other mechanisms without SA-IA-DMA binary and ternary nucleation (Sim_MechNoSA-IA-DMA). January and August were also selected for simulations under the future emission scenario.

Field Measurements.

Field data were measured in January and August 2019 at the Beijing site (BUCT station, 39°56′N, 116°17′E). The SA concentrations were measured by a chemical ionization time-of-flight mass spectrometer (CI-TOF-MS; Aerodyne Research Inc.) (79), and DMA concentrations were recorded by a modified TOF-MS (9, 59). The diethylene glycol–based scanning mobility size spectrometer (DEG-SMPS) was utilized to record the PNSDs spanning a range of approximately 1 nm to 10 µm (80, 81). Additionally, the field data at the Nanjing site were taken from the study of Zhang et al. (28). Further details regarding the observations are available in our previous works (9, 34, 82, 83).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.