Effects of reactive nitrogen gases on the aerosol formation in Beijing from late autumn to early spring

To investigate the formation and evolution mechanism of haze pollution in Beijing cold days, we measured reactive nitrogen gases (e.g. NH3 and HNO3), SO2 and major water-soluble inorganic ions of PM2.5 simultaneously in a two-year (from November to April) study. We found that NH3 and NO3 − have the highest concentrations among the gaseous precursors and inorganic components of PM2.5, respectively. The total NH x (gaseous NH3 and particle NH4 +) was mostly in excess the need to neutralize acid compounds. During the whole study period, the aerosol pH with an average value of 4.05. From normal period into haze episodes, the aerosol pH tends to decrease and the concentration of all species (gases and particles) increases. Meanwhile, declined gas fractions exhibited that enhanced partitioning from HNO3, NH3 and SO2 to their corresponding particle phases. Under the heavy haze period, most HNO3 (79%) has entered into NO3 −, about 41% NH3 remaining as free NH3, while only about 51% of SO2 has been oxidized to SO4 2−, implying the severe Nr pollution in atmosphere of Beijing in winter. Further analysis shows relative humidity (RH) plays an important driving role on the SNA (sulfate (SO4 2−), nitrate (NO3 −), ammonium (NH4 +)) formation and particulate NO3 − formed at a relatively low RH (20%–60%) and SO4 2− at a high RH (40%–80%). Thus, synchronized abatement of multi-pollutants emissions especially for NH3 emission reduction at a regional scale is necessary for mitigating megacities ambient PM2.5 pollution and achieving the UN sustainable development goal through improving N use efficiency in agriculture.


Introduction
With the appearance of United Nations Sustainable Development Goal for 2030, food security and breathing clean air are both important issues worldwide, while managing nitrogen (N) in a sustainable way is crucial to realize both targets (Liu et al 2020b). In recent years, high levels of PM 2.5 have become one of the most serious environmental problems across China and have captured the common attention of the public, governments, and scientists. Studies have shown that secondary inorganic aerosols contribute about 40%-65% of PM 2.5 mass concentration, that are mainly composed of sulfate (SO 4 2− ), nitrate (NO 3 − ), and ammonium (NH 4 + ) (briefly, SNA) (Yang et al 2011, Huang et al 2014, Sun et al 2015. In addition to direct emissions, most SNA are derived from oxidation and neutralization of associated gaseous pollutants, including reactive nitrogen gases (NO x , NH 3 ) and SO 2 (Baek et al 2004, Behera and Sharma 2010, Luo et al 2014. Part of the SO 2 emitted into the atmosphere is oxidized into sulfates and NO x is transformed into nitric acid (HNO 3 ) (Stockwell andCalvert 1983, Seinfeld andPandis 2016). Then, NH 3 neutralizes acid components (SO 2 and HNO 3 ) and reactants dynamically partition into the aerosol phase (Walker et al 2004). Considering PM 2.5 has broad adverse impacts on air quality, climate forcing and human health (Tegen et  China's government has made great efforts on the SO 2 and NO x emissions control since 2005 and resulting in the reduction of 59% and 21% for SO 2 and NO x from 2013 to 2017 (Zheng 2018). Variations in PM 2.5 concentrations are mainly driven by the emission of their precursors (SO 2 and NO x ). The decreased in SO 4 2− corresponded to the changes in SO 2 emissions, whereas the relative increase in NO 3 − was not catchable with NO x emissions reduction (Geng et al 2017. During this time period, little attention has been paid to NH 3 emission control. It has been reported that atmospheric NH 3 experienced an increasing trend or remains stable. Zhang et al (2017) reported total NH 3 emissions substantially increased at an annual rate of 1.9%, from 12.1 Tg N yr −1 in 2000 to 15.6 Tg N yr −1 in 2015 across China. Agricultural activities are the main source of NH 3 , including livestock manure and synthetic fertilizer application, accounting for 80%-90% to total emissions. (Bouwman et al 1997, Xu et al 2019. Whereas there are also other a variety of significant sources, such as fossil fuel consumption, biomass burning and vehicles emission, etc (Behera et al 2013, Meng et al 2017, Chang et al 2019. As a key alkaline species for neutralizing acid pollutants in the atmosphere, several studies revealed the importance of NH 3 to the formation of secondary inorganic and organic aerosols (Babar et al 2017, Meng et al 2018. Taken together, based on the reduction of atmospheric SO 2 emissions in recent years in Beijing, knowledge of the contributions in every precursor especially Nr (NH 3 and NO x ) to PM 2.5 components can be better identifying the air quality issue of Beijing.
Many haze-related aspects of Beijing have been investigated, including PM 2.5 chemical composition (Liu et al 2016, Zhang et al 2018, aerosol formation mechanisms , Duan et al 2019, pollutants regional transport  and effects of meteorology (Dang andLiao 2019, Zhao et al 2019). Despite a large number of aforementioned studies, the evolutionary mechanisms responsible for the gas-particle partitioning during haze events remain poorly characterized, may be due to there were few simultaneous high timeresolution measurements of major precursors and inorganic compositions in PM 2.5 . For example, Geng et al (2017) reported the changes of SO 4 2− and NO 3 − concentrations in PM 2.5 were in line with the increase in NO x emission and the decrease of SO 2 emission while on a national-year scale. Thus, it is worthwhile to study the formation mechanisms for heavy haze via simultaneous measurement of gas precursors and particulate matter, especially in the stage of frequent air pollution. In this study, 2 year daily sampling of HNO 3 , SO 2 , NH 3 and PM 2.5 were conducted at the ground site from November to April in Beijing to fill the gaps in simultaneous measurement. Aerosol acidity and water content were calculated by the thermodynamic equilibrium model (ISORROPIA-II) in order to analyze atmosphere characteristics. Based on the in situ measurements and simulated datasets, we investigated the gas-particle partitioning under different pollution levels and the meteorology effect on SNA formation.

Field measurement site
A Beijing winter field measurement was conducted from November 2017 to April 2018 (first study period, except 2-27 February) and November 2018 to April 2019 (second study period, except 16 January-23 February) at China Agricultural University (CAU, 40 • 01 ′ 29 ′′ N, 116 • 17 ′ 0 ′′ E), which is located outside the north fifth ring road in Beijing. The CAU is a suburban site with mainly influences from local traffic and resident sources. Daily sampling starts at 8:00 am each day and the sampler is replaced at 8:00 am of the next day.

Instrumentation and measure methods
In this study, a University Research Glassware (URG) Corporation annular denuder/filter-pack sampler (URG-3000C, URG Corporation, USA) was deployed to capture both gases and aerosol. First, the URG system draws ambient air through a Teflon-coated PM 2.5 cyclone to remove particles with aerodynamic diameters greater than 2.5 µm. Then, a 10% Na 2 CO 3 solution (water:methanol = 9:1) coated glass denuder captures HNO 3 and SO 2 and a 10% phosphorous acid (water:methanol = 1:9)-coated denuder captures NH 3 . Next, a filter pack with a 37 mm diameter nylon filter captures PM 2.5 . Finally, a supplementary phosphorous acid-coated denuder was placed after the filter to lock any volatilized NH 3 that from particle captured on the filters. Briefly, the URG system actively collects ambient air under the action of the pump at a nominal flow rate of 10 l min −1 . Meanwhile, the sample air volume was recorded by a dry gas meter (Evanoski-Cole et al 2017). After finish sampling, denuder and filter samples were extracted with deionized water (18.2 MΩ) and analyzed by ion chromatography (ICS-600, Dionex, USA). Blank samplers, including field blank and laboratory blank, were analyzed simultaneously.

Data collection
Meteorological data (wind speed, wind direction, temperature, relative humidity (RH)) were collected from the China Meteorological Administration (www.cma.gov.cn/) and weather underground website (www.wunderground.com/). AQI (Air Quality Index) and PM 2.5 data came from China National Environmental Monitoring Centre (www.cnemc.cn/).

Identification of NH 3 -poor and NH 3 -rich
Total NH x concentrations (gas plus particle phases) and required NH x concentrations for overall charge balance were calculated, and excess NH x concentrations to distinguished NH 3 -rich and NH 3 -poor conditions: where, ∑ C i_cations represent the measured nonvolatile cations (Ca 2+ , Mg 2+ , K + , Na + ) needed to be equivalent with all anions ( ∑ C i_anions , including SO 4 2− , NO 3 − , Cl − ) and HNO 3 , HCl in equivalent concentration of hydrogen ion (e.g. for 1 µmol l −1 Ca 2+ = 2 µeq l −1 H + ).

Characterizes of particles formation
Three gas fractions were used to determine the degree of gaseous precursors converted into SNA and indicate secondary transformation processes (Stratton et al 2019): ] .
Similarly, the concentration of all gases and particles are molar concentrations.

Sensitivity of aerosol pH to SO 4 2− , THNO, TNH, RH and T
To explore the major influencing factors on aerosol pH, sensitivity tests were performed (Ding et al 2019).
In the sensitivity analysis, SO 4 2− , THNO, TNH, RH and T were selected as the variables since SO 4 2− , NO 3 − and NH 4 + are major ions in aerosols. To assess how a variable affects PM 2.5 pH, these variables are divided into six levels between measured maximum and minimum, respectively, and then input into ISORROPIA II together with the average of the other species (Na + , Mg 2+ , K + , Ca 2+ , Cl − ). The magnitude of the relative standard deviation (RSD) of the calculated aerosol pH can reflect the impact of variable sensitivity on aerosol acidity. The higher the RSD is, the greater the impact, and vice versa. The average value and variation range for each variable are listed in table S1. The sensitivity analysis in this work only reflected the characteristics during the observation periods, and further work is needed to determine whether the sensitivity analysis is valid in other environmental conditions.

Daily variation of gaseous precursors and particles species
Time series for observed gaseous precursors of HNO 3 , NH 3 , SO 2 and major inorganic aerosol species (pNO 3 − , pNH 4 + , pSO 4 2− ) concentrations were observed in Beijing during November 2017 to April 2018 (except February) and November 2018 to April 2019 (except February) (figure 1). Average active nitrogen and sulfur concentrations were over 1.5 folds greater for the entire second study period compared to the first study period except pSO 4 2− . In general, the average NH 3 concentration was highest (11.7 ± 0.48 µg m −3 ) compared with HNO 3 (6.68 ± 0.29 µg m −3 ) and SO 2 (7.47 ± 0.36 µg m −3 ) and shown monthly temporal variability with higher concentration appeared in early spring, March and April (figure S1). While the concentrations of HNO 3 and SO 2 exhibited an opposite temporal trend that increased in January indicated serious acid gas pollution in Beijing winter. The pNO 3 − concentration in PM 2.5 remained high value with an average of 16.3 ± 1.17 µg m −3 and the ratio of SNA was almost greater than 50%, indicating the need for NO x reductions. In comparison, the pSO 4 2− presented a relatively lower level and averaged 4.39 ± 0.30 µg m −3 during the whole study period. The pNH 4 + concentration peaked in March, while there was no significant difference with other months, and averaged 10.7 ± 0.64 µg m −3 .
The six active nitrogen and sulfur species showed significant correlations (table S2), and higher R square value (R 2 ) between HNO 3 /NO 2 and SO 2 (R 2 = 0.76, 0.52, p < 0.01) suggesting that they may have similar sources or the results of boundary layer movement. The daily average PM 2.5 concentrations of two study periods were 53.5 and 56.0 µg m −3 , both exceeded 35 µg m −3 , the threshold for the Stage I of Chinese National Ambient Air Quality Standard (figure S2), meaning there is still room for air quality improvement. There was good correlation between PM 2.5 and pNO 3 − , pSO 4 2− , pNH 4 + (p < 0.01), respectively. The contribution of SNA to PM 2.5 mass concentration was averaged around 62%, revealing important component of PM 2.5 to be dominated by SNA.

Characteristics of Beijing atmosphere
Excess NH x was calculated to identify NH 3 -poor and NH 3 -rich conditions. The relationship of total NH x and required NH x are shown in figure 2(a). Total NH x was much more than required NH x , indicating that Beijing was characterized by sufficiently free NH 3 . In summary, the amount of total NH x was rich enough to neutralize available acid compounds in gas and particle phases in Beijing from autumn to spring. With the gradual decrease in agricultural activities in Beijing, the agricultural sources of NH 3 emissions are gradually decreasing, especially in winter. In colder days, non-agricultural sources may be the main source of NH 3 . In our previous study, significant high concentrations of NH 3 in air as well as NH 4 + in PM 2.5 at traffic in situ monitoring sites indicate that vehicular source played a vital role in local emission (Xu et al 2017). Isotope technique also reveals persistent non-agricultural and periodic agricultural emissions drive atmospheric NH 3 concentration in urban Beijing (Zhang et al 2020). On the other hand, the large reduction in SO 2 and NO x emissions are limiting particulate matter formation and NH 3 is in large excess (Lachatre et al 2019), especially in North China Plain (Fu et al 2017, Liu et al 2018. Therefore, as SO 2 and NO x emissions continue to decline in the future, there is inevitably going to be more NH 3 free in the atmosphere. Daily aerosol pH and ALWC calculated by ISOR-ROPIA II and displayed in figure 2(b). The pH value  varies in the range of 0.37-7.95 with an average of 4.05 ± 0.09 for the whole period, indicating a medium acidic condition for fine particles. Higher aerosol pH was caused by higher mass fractions of K + , Mg 2+ , Ca 2+ ( figure S3). Overall, aerosol pH during the study periods was comparable to the result (4.2) found by Liu et al (2017a) and that (4.5) found by Guo et al (2017a). The average ALWC was 28.7 ± 4.48 µg m −3 and higher ALWC corresponds to higher PM 2.5 mass concentration (p < 0.01), suggesting an important role in the increase of air pollution. The aerosol hygroscopic growth processes speed up, and the particle surface increases, which enhances the aerosol extinction ability and accelerates the atmospheric chemical reaction rates, favoring the gas to particle heterogeneous reactions (Faust et al 2017).
In turn, it was also reported that ALWC exceeded 200 µg m −3 in winter in Beijing, with nitrate and sulfate playing dominant roles in determining the abundant ALWC (Wu et al 2018) ( figure S4). Generally, pH depends on both the presence of ions and the amount of particle liquid water. There was no correlation between ALWC and pH in this study. This is different from Guo et al (2015): pH pattern was mainly driven by the dilution of aerosol water. Part of (NH 4 ) 2 SO 4 and NH 4 NO 3 will hydrolyze under the high ALWC condition and produce more H + so that a positive correlation between ALWC and pH. Noted that extremely high concentrations of NH x during our study period and produces more portion of NH 4 + , which prevent the dissociation of H + and then hence disturbing the particles acidity (Ge et  , which leads to the maximal changes in aerosol acidity with elevated THNO. There was a relative limit sensitive to pH for TNH. The thermodynamic equilibrium between NH 4 + and NH 3 makes aerosol remain acidic (Weber et al 2016), an increase with a factor of 10 in NH 3 concentrations corresponds to a changing pH by 1 unit (Guo et al 2017a), which may explain the reason that pH is still at low level though there are a lot of NH 3 in Beijing atmosphere. Figures 3(a) and (f) show the gases and particles concentrations at different levels. The AQI is divided into five levels: AQI ⩽ 50 (excellent day), AQI > 50 and ⩽100 (acceptable day), AQI > 100 and ⩽150 (light pollution day), AQI > 150 and ⩽200 (medium pollution day), AQI > 200 (heavy pollution day). 'Acceptable days' was the main air quality level during the study period, indicating a requirement for air quality improvement during autumn to spring. Gas and particle concentrations dramatically increased from clean days to polluted periods, except SO 2 under heavy pollution conditions. On average, the sum of SNA contributed to almost 56% of PM 2.5 during haze episodes (AQI > 100), with NO 3 − , NH 4 + and SO 4 2− , accounting for 30%, 18%, and 8%, respectively. These contributions suggest that NO 3 − plays an important role in the aggravate of pollution level. Meanwhile, the aerosol pH under excellent and good conditions spanned 0-8, while that under polluted conditions was mostly concentrated from 2 to 5. NO 3 − and SO 4 2− mainly concentrated in the fine mode and increased significantly on heavily polluted periods resulted in the growth of acid generation capacity (Pye et al 2020), which may be the main reason for the decrease in aerosol pH consistent with the worsening of air quality level. Additionally, the relationship of gas fractions (FHNO x , FNH x , FSO x ) together with total concentrations of gas and practical (THNO, TNH, TSO) under different pollution levels were illustrated in figures 3(g) and (i). From clean entered into polluted days, THNO, TNH, TSO increased be accompanied by FHNO x , FNH x , FSO x decreased. In other words, less gaseous HNO 3 , NH 3 , SO 2 were related to more portion of NO 3 − , NH 4 + , SO 4 2− in particles with haze aggravated. In the case of heavy pollution, TNH was the highest, up to 4 µmol m 3 , followed by THNO and TSO. Under heavily polluted conditions, about 41% of NH 3 existed in the gas phase. FHNO x and FSO x decreased from 0.53 to 0.21, and 0.77 to 0.49, meaning almost HNO 3 were converted into particle phase, while on the contrary, the oxidation reaction from SO 2 to SO 4 2− was not complete. Similarly, FNO x quantifies the degree of NO 2 and is defined as the mole ratio of NO 2 to total nitrogen (TNO, NO 2 + pNO 3 − ), was shown in figure S5. The TNO increased and FNO x decreased with the aggravation of pollution that has a close agreement with the observed correlation between THNO and FHNO x , which means the transformation degree increases. However, the value of FNO x is higher than that of FSO x , especially when AQI is greater than 100, indicating that the more SO 2 oxidation occurred in the case of serious heavy pollution. This is consistent with the finding by Zhang et al (2018), who also reported the average values of FNO x and FSO x were 0.83 and 0.73 in Beijing winter, respectively. The ratio of total cation to anion was 1.2 on average under serious pollution conditions, meanwhile average ratio between  (b)). During the study period, the concentrations of Cl − ranged from 0 to 0.1 µmol l −1 , donating partly neutralizing effect to cations. In addition, organic acid salts may contribute to charge balance which is not measured in this study. In view of a large amount of NO x in the atmosphere, the neutralization potential of NH 3 (text S1) was calculated in figure S6, and it was found that there was still a strong neutralization room for HNO 3 even under the serious haze.

Influence of meteorology
A similar time of two samplings in this study avoided a bias from seasonal differences. The temperature varied slightly between the two study periods (p > 0.05), with the first period being −9.71 • C to 21.6 • C which is colder on average than another study period (table  S3). Higher temperatures will facilitate the volatility of NH 3 (p < 0.05) and result in higher NH 3 concentrations in early spring than in winter (Burch and Fox 1989), suggests a higher risk of NH 3 emissions from volatility-driven sources in warm days. Different from NH 3 , there were negative correlations between HNO 3 , SO 2 concentrations and temperature during the observation period. Wintertime home heating in Beijing and surrounding provinces (November-March) is fully covered in this study. At present, natural gas/electric heating contributes almost 97% in Beijing and gradually spread in the surrounding provinces, while coal-fired heating still remains dominant in north China (http://tjj.beijing.gov.cn/). Natural gas consumption could account for 23.4% of total NO x emissions in the heating season in Beijing (Xue et al 2017). At the same time, the lower mixed layer height in winter could lead to pollutant accumulation and adverse to dispersion.
Previous studies have shown a strong correlation between regional transport patterns and local air quality in Beijing, with wind speed and direction playing an important role , Duan et al 2019. However, the concentrations of gas precursors and corresponding PM 2.5 ion components were higher in the second study period compared to the first one, but the winds in the second period were more frequent and stronger from the north, which was thought to deliver more clean air masses (figures S7 and S8). Figures 4(a) and (b) shows the 24 h back trajectories of the air masses with the highest 10% of THNO (HNO 3 + pNO 3 − ), TNH (NH 3 + pNH 4 + ), TSO (SO 2 + pSO 4 2− ) concentrations in Beijing. Back trajectories clearly showed that the highest concentrations associated with transport mainly from the south, and the significant contribution from Hebei Province for the second period about 45.8%. Noncentralized coal heating, un-fully controlled sources in heavy industries, intensive agriculture and dominated highway transportation have caused serious pollutant emissions in Hebei province, which in turn has aggravated Beijing air quality through regional transport . Figures 4(c) and (d) present gas and particle concentrations at different RH and wind directions, which coincides with simulation of backward trajectory and further explains the difference between our study and normal cognition. Overall concentrations were closely related to RH. Air masses from the south (Hebei province) during the second period were wetter and may carry more anthropogenic gas and particulate matter, or gas precursors underwent rapid oxidation in the humid atmosphere environment of Beijing, which had led to a higher sensitivity of aerosols regional transport to Beijing even under the relatively lower frequency of prevailing south winds.
As shown in figure S9, higher RH were generally associated with polluted days than clean days, with RH higher than 57% in pollution days that AQI > 100 and averaged 36% during clean days (AQI = <100). There were positive correlations between RH and FHNO x , FNH x , FSO x , indicating the contribution of aqueous-phase processing to the SNA formation. Apart from providing the available heterogeneous reacting medium via enhancing aerosol surface areas and volumes, liquid water also can liquefy the aerosol particles, increase the ALWC and uptake due to solubility, and reduce the kinetic limitation of mass transfer for gaseous precursors (Engelhart et al 2011, Xie et al 2017. As clearly shown in figure 5, the more complete conversion of HNO 3 to pNO 3 − at the lower RH, which displays formation of NH 4 NO 3 occur at RH from 20% to 60% under NH 3 -rich atmosphere in our case, consistent with the previous study (Quan et al 2015). NO 3 − is formed from NO x through the gas-phase reaction between OH and NO 2 during the daytime and the heterogeneous uptake of N 2 O 5 during the nighttime (Baasandorj et al 2017, Yun et al 2018. The hygroscopic growth of particles facilitated the condensational loss of N 2 O 5 and HNO 3 to particles contributing pNO 3 − . In this study, day and night samples of gas and particulate matter from a pollution event were collected. The ambient RH increased from 53% (daytime) up to 78% (nighttime) during 22 April 2019 and 38% (daytime) up to 73% (nighttime) during 25 April 2019 when pNO 3 − concentrations increased by about twice (figure S10), suggesting nocturnal pNO 3 − formation maybe stronger than that of daytime during the haze period. Different from HNO 3 , rapid formation of SO 4 2− occurs at RH between the range of 40% and 80%, as estimated by about 80% by Kreidenweis and Asaawuku (2014). Generally, SO 2 is oxidized to pSO 4 2− via reaction with OH or aqueous reactions driven by O 3 and H 2 O 2 or transition metal ions. SO 2 oxidation by NO 2 has recently been proposed as an important oxidant in urban areas (Cheng et al 2016, Au Yang et al 2018. Neglecting the NO 2 oxidation pathway leads to overestimate SO 2 oxidation by transition metal ions and OH. In fact, SO 2 oxidation rate is very sensitive to both pH and oxidant concentration. According to Guo et al (2017), major SO 4 2− oxidation through a NO 2 -mediated pathway is not likely in China where the aerosol is consistently more acidic. Actually, NO 2 acted as the initiator of radical formation and the synergy of NO 2 and O 2 resulted in much fast SO 4 2− formation . The heterogeneous reactions of atmospheric oxidants with liquid core of aerosols can be also influenced by ionic strength, which in turn can impact important atmospheric processes. For example, high solute strength of the aerosol particles significantly enhances the SO 4 2− formation rate for the H 2 O 2 oxidation pathway (Liu et al 2020a, Mekic andGligorovski 2021). Furthermore, recent studies have shown aqueous-phase SO 2 oxidation and SO 4 2− formation by NO 2 and OH produced from photolysis of pNO 3 − with high RH and NH 3 neutralization or under cloud conditions , Gen et al 2019. It is worth noting that the above-mentioned oxidation pathways were under abundant liquid water conditions. Thus, an effective SO 2 oxidation mechanism requires higher RH. Gas fractions decreased with increasing RH likely due to these aqueous reactions. Most importantly, once polluted gases are emitted or transported, combined with moist air masses, the positive feedback 'RH-ALWCheterogeneous reactions' would be triggered: ALWC rises as RH increases, and provides the medium for heterogeneous reactions that will result in the secondary aerosol formation, which in turn will increase ALWC by the contribution of hygroscopic accumulation mode particles (Tan et al 2017.

Conclusions
In this paper, we simultaneously measured daily precursor gases (HNO 3 , NH 3 , SO 2 ) and PM 2.5 during November 2017 to April 2018 and November 2018 to April 2019 (excluding February in both periods) in Beijing to investigate the temporal variation, atmospheric characteristics, gas-particle partitioning and meteorological influencing factors.
These results clearly indicated total NH x was mostly in excess of the SO 4 2− -NO 3 − -NH 4 + -water equilibrium system and it is necessary to identify the complex urban NH 3 sources and joint control with acid pollutants. Surprisingly, aerosols in Beijing (winter) are still medium acidic. Compared with the clean days, the polluted episodes were characterized as having more gas and particles, lower pH value and higher ALWC. Under heavy polluted condition, almost HNO 3 were entering into particle phase, while the oxidation reaction of SO 2 was not complete. Further analysis revealed that RH is an important driving factor in haze evolution. Results found that the formation of NH 4 NO 3 occurred under RH < 60%, while SO 4 2− increased sharply in the state of RH around 40%-80%. Although our finding has been conducted by analyzing observations in Beijing, it may be also applicable to other mega cities with similar emissions and meteorological conditions. Meanwhile, it is necessary to further explore the differences of gasparticle partitioning in urban and rural areas. During the hazy periods of Beijing in winter, about 40% of NH 3 still remains in atmosphere as the gas. Considering the significant effect of aerosol acidity on secondary aerosol formation and the solution of metals by acid dissociation, as well as gradually mitigated SO 2 pollution, the emissions of Nr need more attention, especially for NH 3 since NO x concentrations have recently been decreasing with years (Wen et al 2020). In addition, in view of the role of transportation, regional joint control and prevention are essential. For example, Hebei is a province with intensive agricultural activities close to Beijing that produces large amounts of NH 3 and contributes to the formation of aerosols, but these aerosols and some precursors may be transported to Beijing, causing local haze pollution. Therefore, an important work of the next stage in China is to achieve the UN sustainable development goal through four improved nitrogen management strategies: improved farm management practices with nitrogen use reductions, machine deep placement of fertilizer, enhanced-efficiency fertilizer use, and improved manure management (Guo et al 2020).

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.