The Synergistic Role of Sulfuric Acid, Bases, and Oxidized Organics Governing New‐Particle Formation in Beijing

Intense and frequent new particle formation (NPF) events have been observed in polluted urban environments, yet the dominant mechanisms are still under debate. To understand the key species and governing processes of NPF in polluted urban environments, we conducted comprehensive measurements in downtown Beijing during January–March, 2018. We performed detailed analyses on sulfuric acid cluster composition and budget, as well as the chemical and physical properties of oxidized organic molecules (OOMs). Our results demonstrate that the fast clustering of sulfuric acid (H2SO4) and base molecules triggered the NPF events, and OOMs further helped grow the newly formed particles toward climate‐ and health‐relevant sizes. This synergistic role of H2SO4, base species, and OOMs in NPF is likely representative of polluted urban environments where abundant H2SO4 and base species usually co‐exist, and OOMs are with moderately low volatility when produced under high NOx concentrations.


Introduction
Atmospheric aerosol particles substantially affect climate and human health (Heal et al., 2012). Most aerosol particles in the Earth's atmosphere, including a large fraction of cloud condensation nuclei (CCN), are thought to originate from formation of atmospheric clusters and their subsequent growth to larger sizes (Gordon et al., 2017;Merikanto et al., 2009;Spracklen et al., 2008). This process is known as new particle formation (NPF). NPF occurs in almost all continental environments with varying frequencies and intensities (Chen et al., 2012;Kalafut-Pettibone et al., 2011;Kerminen et al., 2018;Kulmala et al., 2017;Mönkkönen et al., 2005;Nieminen et al., 2018;Peng et al., 2014;Qi et al., 2015;Wu et al., 2007;Xiao et al., 2015;Yao et al., 2018), and therefore, understanding the underlying mechanisms is an on-going research topic of high importance.
In spite of high primary aerosol emissions, NPF is a major source of ultrafine particles in polluted urban environments (Guo et al., 2014;Kulmala et al., 2016;Wang, Wu, Yue, et al., 2017). Studies have also suggested that NPF in urban environments contributes more to the CCN budget than it does in regional or remote environments Wang, Wu, Yue, et al., 2017). In addition, NPF may also lead to haze under stagnant air conditions (Guo et al., 2014;Kulmala et al., 2021). Previous studies have suggested that high aerosol surface concentration is one primary parameter that governs the occurrence of NPF in polluted environments Deng et al., 2021;McMurry et al., 2005). Another recent study has predicted that little NPF should occur in polluted megacities due to the high condensation sink (CS) (Kulmala et al., 2017); by contrast, observations show that NPF in these areas are frequent and more intense than in clean environments (Deng et al., 2020). This discrepancy underlines an incomplete understanding of NPF in polluted and usually chemically complex environments.
Up to now, the driving mechanisms and key contributing species in both atmospheric clustering and subsequent growth in urban environments remain to be elucidated. For instance, a recent study has shown that clustering of gaseous sulfuric acid (H 2 SO 4 ) and dimethylamine (C 2 H 7 N) is responsible for the initial steps of NPF in Shanghai . However, another recent laboratory-based study suggested that low-volatility organic vapors from the oxidation of vehicle-emitted organics, rather than H 2 SO 4 , can initiate the NPF in some urban environments (Guo et al., 2020). In addition, it was recently suggested that various dicarboxylic acids (diacids) could also cluster with H 2 SO 4 and C 2 H 7 N to initiate NPF (Fang et al., 2020).
Moreover, the contribution of low-volatility organic vapors to the NPF in the polluted urban atmosphere is largely unclear. These vapors form via the atmospheric oxidation of various volatile organic compounds (VOCs). It is widely acknowledged that low-volatility organic vapors are likely dominant contributors to the growth of freshly formed molecular clusters, allowing them to survive the scavenging in the atmosphere (e.g., Riipinen et al., 2011;Smith et al., 2008). Such contribution is controlled by both the concentration and saturation vapor pressures of low-volatility organic vapors (Donahue et al., 2013). Extensive studies that have been conducted in remote forested areas demonstrated that organic vapors, formed via the oxidation of monoterpenes, are able to contribute to the particle growth below 2 nm (Kulmala et al., 2013;Mohr et al., 2017). By contrast, the chemical characteristics of these low-volatility vapors in urban settings are under-studied, and the potential contribution of these vapors to NPF remains to be evaluated.
To advance our understanding of NPF in urban environments, we built a station in downtown Beijing equipped with state-of-the-art instruments. Our measurements cover most of atmospheric species directly contributing to NPF and particle evolution over a wide diameter range (1 nm-10 μm). Our comprehensive measurements allow for molecular-level analyses on NPF, upon which we determine the critical NPF processes and demonstrate the roles of different species in these processes.

Measurement Station
The measurement was conducted at the Aerosol and Haze Laboratory/Beijing University of Chemical Technology station (AHL/BUCT station). The station is located in the main teaching building of the west cam-YAN ET AL. 10.1029/2020GL091944 2 of 12 contribution to the formation of new particles but were crucial for particle growth above 2-3 nm to climate-and health-relevant sizes.
pus of Beijing University of Chemical Technology (Lat. 39º56ʹ31ʺ and Lon. 116º17ʹ52ʺ). The building is surrounded residential and commercial areas. In addition, there are three main roads within 600 m around the station. Therefore, the AHL/BUCT Station is a typical urban observation station affected by heavy traffic and residential emissions. More details about the station can be found elsewhere (Liu et al., 2020).

Instrumentation
The particle number concentration and size distribution over in the diameter range of 1 nm-10 μm were measured by a diethylene glycol scanning mobility particle spectrometer (DEG-SMPS, 1-7.5 nm) and a particle size distribution system (3 nm-10 μm). Additionally, the number concentration of clusters with a diameter range of 1.3-2.5 nm was measured with a Particle Sizer Magnifier, which showed good agreement with the DEG-SMPS measurements. The particle formation rate (J 1.5 ) was calculated for all NPF cases based on these data sets using the new-balance formula optimized for polluted environments . H 2 SO 4 and oxidized organic molecule (OOM) concentrations were measured with nitrate ion-based Chemical Ionization Atmospheric Pressure interface Time-of-Flight mass spectrometer (CI-APi-TOF, Aerodyne Research, Inc.). This instrument was calibrated for both the sensitivity to H 2 SO 4 and the mass-dependent transmission efficiency. The molecular ion cluster composition was measured with an Atmospheric Pressure interface Time-of-Flight mass spectrometer (APi-TOF, Aerodyne Research, Inc.). The OOM volatility was estimated based on the parameterization by Mohr et al. (2019) and further corrected using the ambient temperature. More details of these instruments and other ancillary measurements are provided in the supporting information.

The Frequent Occurrence of NPF Events and the Governing Factors
We extend the mechanistic understanding of NPF in wintertime Beijing based on the results from intensive measurements conducted during January 23-March 31, 2018 at the newly built station in downtown Beijing. In Figure S1, we present the measurement overview, including the particle number concentration and size distribution over a wide size-range from 1 nm to 10 μm, relevant meteorological variables (temperature and UVB radiation intensity), calculated particle formation rate at 1.5 nm (J 1.5 ), CS, as well as gas-phase concentrations of vapors plausibly contributing to NPF (i.e., H 2 SO 4 and OOMs). Details of these measurements and calculations are provided in the supporting information. We classify the whole measurement period into NPF and nonNPF days following the widely used method suggested in previous studies (Dal Maso et al., 2005). During the measurements, we observed NPF events frequently (32 out of 69 days), and during the first five weeks from 23 January to 28 February, NPF occurred on more than 60% of the days. The reduced NPF frequency in March is clearly associated with high CS, which is usually encountered when air masses come from polluted areas located on the south and east of Beijing. This is also consistent with an earlier study conducted at another site in Beijing .
Comparisons of different variables between NPF and nonNPF days are shown in Figure 1. In the present-day continental boundary layer, it is widely accepted that H 2 SO 4 and low-volatility organic species are crucial species for NPF Riccobono et al., 2012Riccobono et al., , 2014Rose et al., 2018;Yao et al., 2018). However, both H 2 SO 4 monomer and total OOM concentrations in our observations were higher on nonNPF days than on NPF days (Figures 1b and 1d), likely due to the higher precursor concentrations (SO 2 and VOCs). Note that this observation should not be interpreted as indicating that H 2 SO 4 and OOMs do not participate in NPF, but rather that enhanced clustering can be outcompeted by the simultaneously elevated loss of clusters to the condensation sink (i.e., higher CS, Figure 1e). The competition between H 2 SO 4 clustering and CS is reflected in the H 2 SO 4 dimer concentration, which is lower on nonNPF days due to the enhanced scavenging ( Figure 1c). The close association between H 2 SO 4 dimer and NPF occurrence and intensity will be discussed in detail in the next section. Furthermore, the temperature can influence NPF by modifying the formation and volatility of OOMs (Simon et al., 2020;Stolzenburg et al., 2018) and affecting the stability of H 2 SO 4 clusters . However, the mean temperature in NPF and nonNPF days was almost identical (Figure 1f), suggesting that temperature does not determine the NPF occurrence during our measurement period.

Intense New Particle Formation Driven by Acid-Base Clustering
In addition to the high frequency, NPF also occurred with high intensities, with a median J 1.5 of 86 cm −3 s −1 at peak hours ( Figure 1a). The high particle formation rates appear to be a common characteristic of NPF in the polluted urban atmosphere, such as Shanghai (Xiao et al., 2015;Yao et al., 2018), Nanjing (Yu et al., 2016), and Barcelona (Brean et al., 2020). These particle formation rates are usually 1-2 orders of magnitude higher than those in relatively clean environments (Kulmala et al., 2013), suggesting the existence of abundant and efficient nucleation species in an urban atmosphere.
In Figure 2a, we show how the particle formation rate J 1.5 changes as a function of H 2 SO 4 concentration. Despite an overall weak correlation between J 1.5 and H 2 SO 4 monomer concentration, strong correlation becomes visible within narrow ranges of CS. This suggests that while H 2 SO 4 drives the initial steps of NPF (i.e., the formation of embryonic clusters), CS plays a crucial role in determining the NPF intensity. For instance, at the same H 2 SO 4 monomer concentration, J 1.5 may vary up to 3-4 orders of magnitude when CS ranges from 0.001 to 0.1 s −1 . Therefore, such J 1.5 -H 2 SO 4 relationship should be interpreted with caution when inferring the NPF mechanism, especially when comparing relatively more polluted ambient data with chamber results, because chambers usually have a lower CS . In our case, the J 1.5 -H 2 SO 4 relationship at CS < 0.02 s −1 agrees well with the results of the CLOUD chamber experiments using YAN ET AL.
10.1029/2020GL091944 4 of 12 Figure 1. Diurnal variations of particle formation rates as well as relevant species and parameters in new particle formation (NPF) and nonNPF days. Lines are the median values, and shaded areas denote the 25%-75% of the data variation.
H 2 SO 4 and dimethylamine (C 2 H 7 N) as the nucleating vapors (Kürten et al., 2018), suggesting that H 2 SO 4 clustering that proceeds at similar efficiency is driving the initial steps of NPF.
To further confirm the high efficiency of H 2 SO 4 cluster formation, we investigate the birth-death processes of H 2 SO 4 dimers, that is, clusters containing two H 2 SO 4 and also base molecules, which are the only H 2 SO 4 clusters that can be reliably quantified. Other larger clusters also form, but they usually fall below the instrumental detection limit due to their low concentration and reduced ionization efficiency (thus low sensitivity) (Jen et al., 2016). As shown in Figure 2b, we found that the H 2 SO 4 dimer concentration was in a pseudo steady-state between its formation rate by H 2 SO 4 monomer collisions and its loss rate onto pre-existing larger particles by coagulation scavenging. The second-order rate constant between H 2 SO 4 monomer collision forming one H 2 SO 4 dimer is determined as 3.89 × 10 −10 cm 3 s −1 (the fitted slope in Figure 2b), which is close to the theoretical molecular collision rate constant. Moreover, we observe a very strong correlation (R 2 = 0.82) between J 1.5 and H 2 SO 4 dimer concentration (Figure 2c), showing that the abundance of H 2 SO 4 clusters is directly associated with NPF intensity.
To further understand the high efficiency of H 2 SO 4 clustering, we characterize the composition of H 2 SO 4 cluster ions measured by the APi-TOF during the NPF events. As shown in Figure 2d, in addition to clusters consisting purely of H 2 SO 4 (red dots in Figure 2d), amines (C 2 H 7 N, C 3 H 9 N, C 4 H 11 N) start to appear from the H 2 SO 4 trimer onwards (magenta and blue dots in Figure 2d). The appearance of C 2 H 7 N in the H 2 SO 4 trimer is consistent with the results in the CLOUD chamber . In addition, C 3 H 9 N and C 4 H 11 N participate in H 2 SO 4 clustering at the same step; consistent with an earlier laboratory study with different amines (Jen et al., 2014), this suggests that C 3 H 9 N and C 4 H 11 N have a similar stabilizing effect on H 2 SO 4 clusters. In fact, these amines form stable neutral clusters with one or two H 2 SO 4 . However, as suggested by our quantum chemical calculations using C 2 H 7 N as a surrogate, deprotonation of these clusters will result in the loss of the amine moieties (see supporting information Section 3). Thus, large fractions of HSO 4 − or H 2 SO 4 HSO 4 − signals in our mass spectra come from (H 2 SO 4 ) 1,2 −(C 2-4 H 7-11 N) 1,2 clusters. Other base molecules, such as CH 5 N and NH 3 are involved in the cluster formation at later steps, as they are observed only in larger clusters. Moreover, unlike observations in forested areas and chamber experiments with the presence of both H 2 SO 4 and biogenic VOCs Riccobono et al., 2014;Schobesberger et al., 2013), H 2 SO 4 clusters containing both base molecules and oxygenated organic molecules are not observed, suggesting little importance of multi-component nucleation. Altogether, we conclude that the initial steps of NPF, that is, the formation of 1.5 nm particles, is predominantly driven by acid-base clustering during the measurement period.

Characteristics of Oxygenated Organic Molecules and Their Role in NPF
In order to evaluate the role of low-volatility organic vapors in NPF, we measured organic molecules with the nitrate anion CI-APi-TOF (see more details in the supporting information). As we will demonstrate below, according to the latest definition of highly oxygenated organic molecules (HOMs) (Bianchi et al., 2019), most organic vapors that we detected in Beijing do not meet that requirement. Therefore, we use the term "oxygenated organic molecules (OOMs)" to refer to the measured compounds in this study. One should note that this instrument does not detect all OOMs, but it is specifically selective toward more oxidized ones, which are also most relevant for NPF (Ehn et al., 2014). In addition, we found this instrument to be sensitive to various nitrated phenols. However, these nitrated phenols are too volatile to have any noticeable contribution to NPF and are therefore excluded in this study. All identified OOMs together with their concentrations are listed in Table S2.
In Figure 3, we show chemical characteristics of OOMs based on the number of carbon atoms (nC), numbers of nitrogen atoms (nN), and effective oxygen atoms (nO eff ). The nO eff is calculated by disregarding the two oxygen atoms bonding with each nitrogen atom to reflect better the oxidation state and volatility of OOMs . These characteristics provide important information for understanding the critical processes in OOM formation.
Our first observation is that OOMs form through the oxidation of a complex mixture of VOCs. As suggested by the carbon distribution of OOMs that C 5-10 OOMs contribute to a dominant fraction of the total OOM concentration (Figure 3a), isoprene (C 5 ), alkylbenzenes (C 6-9 ), and monoterpenes (C 10 ) are likely the most important precursor VOCs. In addition, C 3-4 OOMs might arise from the decomposition during the oxidation of VOCs with larger carbon numbers, and the C >10 may come from the oxidation of polyaromatic hydrocarbons or from accretion reactions between two peroxy radicals (RO 2 ). However, it should be noted that, not all identified peaks in this study (Table S2) can be found in previous laboratory studies (Ehn et al., 2014;Krechmer et al., 2015;Molteni et al., 2018;Wang, Wu, Berndt et al., 2017, Wang et al., 2018. The difference in OOM composition may arise from the imperfect mimicking of the VOC oxidation under real atmospheric conditions in previous laboratory experiments, and also possibly results from some important but uncounted OOM sources, such as missing precursor VOCs and the evaporation from particulate phase.
YAN ET AL.
10.1029/2020GL091944 6 of 12 Our second observation is that a major fraction (∼75%) of OOMs are organic nitrates (Figure 3b), suggesting that the reaction of RO 2 + NO x is the main termination reaction pathway. In Beijing, the concentrations of peroxy radicals (HO 2 and RO 2 ) and NO x are in the ranges of 10 7 -10 8 cm −3 and 10 11 -10 12 cm −3 , respectively (Tan et al., 2018). Given the fact that the reaction rate coefficient between RO 2 and HO 2 /RO 2 is at most one order of magnitude higher than that between RO 2 and NO, the reactions of RO 2 + RO 2 or RO 2 + HO 2 are less important channels than the RO 2 + NO x channel in OOM formation. In addition to terminating the oxidation, the reaction of RO 2 + NO can also propagate the oxidation by forming alkoxy radicals (RO), which does not lead to the formation of a nitrate group. As a result, it is likely that the actual contribution of the RO 2 + NO x reaction to the formation of OOMs is even greater than 75%. Furthermore, about 30% of OOMs contains two or three nitrogen atoms ( Figure 3b). As indicated by the higher concentration during the day than at the night ( Figure S2), the formation of such multiple nitrate groups was associated more with photochemistry than with the dark NO 3 chemistry. Multiple oxidation steps each followed by a RO 2 + NO x termination reaction could plausibly explain the formation of two or three nitrogen-containing compounds during daytime.
Our third observation is that HOMs in our measurements constituted a minor fraction of all OOMs. This could be explained by high NOx concentrations, which generally suppress RO 2 autoxidation (Figure 3c). From the literature, it is known that the degree of suppression is highly dependent on the precursor VOCs. For isoprene, 5 ppb NO is sufficient to completely shut down the autoxidation at 298 K (Wang et al., 2018). Due to the lower temperature during our measurement, which further slows down autoxidation, and higher NO x concentration, autoxidation in isoprene oxidation should be fully prevented. For monoterpenes, autoxidation should also be significantly suppressed. This is supported by the comparison of the C 10 H 15 O x N product distribution in Beijing and a boreal forest in southern Finland where the NO x concentration is much lower . As shown in Figure S3, relatively less oxygenated products (C 10 H 15 O 6-8 N, nO eff = 4-6) are the main products, while highly oxygenated products (C 10 H 15 O 8-11 N, nO eff = 6-9) at the boreal forest are dominant ones. For alkylbenzenes, the autoxidation rate of the bicyclic peroxy radicals (BPRs) varies by three orders of magnitude depending to the alkyl substitution group (Wang, Wu, Berndt, et al., 2017). The suggested maximum autoxidation rates of isopropyl-benzene, ethyl-benzene, toluene at 278 K are 9.2, 4.6, and 0.01 s −1 , respectively (Wang, Wu, Berndt, et al., 2017). These rates are equivalent to the bimolecular reaction rate with NO of about 38, 18, and 0.04 ppb, respectively, assuming a rate constant of 1 × 10 −11 cm 3 s −1 . Therefore, the autoxidation for alkylbenzenes with long-chain substituents (e.g., isopropyl-, ethyl-benzenes) should only be partially suppressed in Beijing, in contrast to the complete suppression of autoxidation for alkylbenzenes with short-chain substituents (e.g., methyl-, dimethyl-benzenes).
The volatility of OOMs is a key variable, linking the formation and chemical composition of OOMs to their potential contribution to the particle growth. We estimate the volatility distribution of the OOMs using a volatility basis set parameterization (see Section 2) (  binned volatility distribution is shown in Figure 4a. Based on our estimate, ultralow volatility organic compounds (ULVOC, C* ≤ 3 × 10 −9 μg m −3 ) that are capable of nucleating (Schervish & Donahue, 2020) are minor products, with a mean concentration of 2 × 10 5 cm −3 (0.7% of total OOMs). This observation is also consistent with our former conclusion that the main NPF-relevant clustering process is H 2 SO 4 -base that does not involve OOMs. Also, only a small fraction of OOMs (9%, corresponding to an absolute concentration of 3 × 10 6 cm −3 ) is extremely low-volatility organic compounds (ELVOC, 3 × 10 −9 < C* ≤ 1 × 10 −4.5 μg m −3 ). Such low concen-YAN ET AL.
10.1029/2020GL091944 8 of 12 trations suggest that organic vapors are not the main driver of particle growth at the smallest size (e.g., ∼2 nm). Consistent with this, an independent study based on H 2 SO 4 measurement also indicates that H 2 SO 4 and its clusters contribute significantly to the particle growth at 1.5-3 nm (Deng et al., 2020). ULVOCs are almost exclusively dimer compounds formed via association reactions between RO 2 radicals, while ELVOCs include both dimer compounds and highly oxygenated monomer compounds. Therefore, their low concentrations can be explained by the dominance of RO 2 +NO x as the termination reaction, as we have discussed above.
Even though most OOMs (59%) in Beijing are semi-volatile organic compounds (SVOCs, 10 −0.5 < C* ≤ 10 2.5 μg m −3 ) and intermediate volatile organic compounds (IVOC, 10 2.5 < C* ≤ 10 6.5 μg m −3 ) that do not drive the growth of newly formed particles, a considerable fraction (31%) are low-volatility organic compounds (LVOCs, 10 −4.5 < C* ≤ 10 0.5 μg m −3 ) that can contribute to the growth of particles above ∼2 nm. This contribution is important for the survival of new particles against the scavenging loss. On NPF days, when newly formed particles grew beyond 30 nm, the concentration of OOMs was significantly higher than on days when particles did not reach 30 nm (Figure 4b). This suggests that condensable OOMs are crucial for particle growth from a few nanometers to a few tens of nanometers. It is worth noting that, the OOM volatility distribution in both cases were similar, indicating that the difference of OOMs was mainly caused by the overall VOC oxidation rate rather than a significant change in RO 2 + NO x chemistry, which can alter the OOM volatility distribution but not the overall concentration .

Conclusion
In summary, with state-of-the-art instruments, we track the particle evolution from molecules to clusters and then to particles. Based on such comprehensive measurements, we depict a complete picture of NPF in wintertime Beijing, including both the initial clustering and the subsequent cluster growth driven by different vapors. The clustering starts with the formation of thermodynamically stable sulfuric acid clusters at around 1-1.5 nm. This process in wintertime Beijing is driven by the acid-base clustering mechanism in the presence of abundant stabilizers, such as amines and NH 3 . This is confirmed by our observations from many independent perspectives, including the J-H 2 SO 4 -CS relationship, the direct observation of molecular ion clusters by APi-TOF, and the low concentrations of ULVOC and ELVOC. This is in contrast to a recent suggestion made based on chamber experimental results that OOMs are solely responsible for NPF in Beijing. Nevertheless, OOMs are crucial for growing the newly formed particles through the very small sizes where they are vulnerable to scavenging loss. We confirm this, by quantifying each OOMs volatility classes. Overall, we clearly demonstrate that H 2 SO 4 , various bases species, and low-volatility organic vapors synergistically contributed to the intense NPF in wintertime Beijing, and in order to control the aerosol population, those key compounds need to be monitored and regulated.