Severe haze pollution has been causing serious environmental problems and harming public health in China over the past decades (He et al., 2001; Wang et al., 2016; R. Zhang et al., 2015). Therefore, strong actions have been taken to improve the worsening atmospheric environment, including cutting down the pollutant emissions with forced installation of catalytic converters on vehicles, building clean-coal power generation systems, prohibiting open burning of crop residue during the harvest seasons, etc. (Chen et al., 2017; Zhang et al., 2012; Liu et al., 2016). As a result, the PM2.5 pollution occurrence has reduced to meet the goals in the Air Pollution Prevention and Control Action Plan (issued by the State Council of China, http://www.gov.cn/zwgk/2013-09/12/content_2486773.htm, last access: 24 February 2020, in Chinese). Among all the regions of interest, Beijing has achieved great success in PM2.5 reduction (the annual average PM2.5 concentration of 2017 was 58 µgm-3). Yet, PM2.5 concentration in Beijing is still higher than that in most developed countries.

There are many factors contributing to the PM2.5 pollution in China (Guo et al., 2014; Ding et al., 2013). The PM2.5 pollution across the country is characterized by significantly high secondary formation of inorganic components (R.-J. Huang et al., 2014). Sulfate, nitrate, and ammonium (SNA) comprised over 30 % of the PM2.5 mass, and SNA's fraction continues to increase during pollution evolution (Cao et al., 2012). While models can predict the airborne particle pollution in the US or Europe well, it is challenging to simulate the real atmospheric pollution in China (Wang et al., 2014; Ervens et al., 2003). Previous modeling studies showed that the simulated PM2.5 concentrations were underestimated within the current scheme, which is related to the important role of heterogeneous reactions in SNA formation processes (X. Huang et al., 2014; Herrmann et al., 2005). It was reported that the classical formation mechanism of sulfate in the atmosphere was through oxidation by H2O2. But recent studies in China pointed out that nonclassical formation pathways cannot be ignored. In Beijing, severe haze events occur with abundant nitrogen species (NOx, NH3, etc.), high relative humidity (RH), and less active photochemistry (Wang et al., 2016; Cheng et al., 2016). Field observations, chamber experiments, source apportionments, and numerical simulations all suggest that the joint effect of NO2, SO2, and NH3 is important for sulfate formation processes in haze events (Cheng et al., 2016; Wang et al., 2016; G. Wang et al., 2018; He et al., 2018, Xue et al., 2019). Aqueous oxidation of SO2 by NO2 as well as the catalyzed oxidation by transition metal ions (TMIs) could be major processes of sulfate formation in Beijing during winter (Wang et al., 2016; Cheng et al., 2016). In addition, although the photochemistry is less active during haze periods in winter, the extra OH radical provided by HONO might enhance the atmospheric oxidation capacity and lead to rapid formation of SNA (Tan et al., 2018, Ge et al., 2019). Since these reactions are all sensitive to particle acidity, adequate quantification of airborne particles' acidity is essential for elucidating the specific contribution.

Particle acidity has been widely studied due to its important role in haze formation and has been widely implemented in major models (Yu et al., 2005; Oleniacz et al., 2016). Since the practical method of directly measuring the particle acidity in the real atmosphere is not available (Wei et al., 2018; Freedman et al., 2019), thermodynamic models have been mostly used in quantifying the particle acidity. Most models (ISORROPIA II, E-AIM-IV, AIOMFAC, etc.) can predict H+, aerosol liquid water content (ALWC), and the partitioning of volatile and semi-volatile components, such as ammonia (Fountoukis and Nenes, 2007; Clegg et al., 2008). These models' abilities to describe physiochemical properties of airborne particles have been validated in previous studies (Weber et al., 2016; Guo et al., 2016; Shi et al., 2017; Tao and Murphy, 2019; Murphy et al., 2017). However, several publications using the same method gave different particle pH values in Beijing, and contradictory conclusions were drawn on the importance of sulfate formation by NO2 oxidation. Cheng et al. (2016) conducted some modeling work and suggested that the PM2.5 pH in Beijing ranges between 5.4 and 6.2, which is favorable for the aqueous NO2 oxidation. The NO2 oxidation's major contribution and the importance of high ALWC and sufficient ammonia are supported not only by modeling works, but also by field observation and chamber studies (Wang et al., 2016; Chen et al., 2019). Conversely, Liu et al. (2017) simulated the particle pH during the winter of 2015 and 2016 with the same method and claimed that pH of the Beijing haze particles was lower (3.0–4.9, average 4.2) and unfavorable for the NO2 oxidation mechanism. Based on the ISORROPIA II model results, which assumed Chinese haze particles to be a homogeneous inorganic mixture, Guo et al. (2017) further concluded that high ammonia cannot increase the particle pH enough for the aqueous oxidation of SO2 by NO2. Recently Song et al. (2018) reported that the thermodynamic model (ISORROPIA II) has coding errors, which can lead to the predicted pH values being negative or above 7. Furthermore, with lab studies and field observations, G. Wang et al. (2018) raised the concern of whether it is appropriate to elucidate sulfate production for Beijing haze by using particle pH predicted only based on the inorganic compositions. In fact, since the real atmosphere is affected by uncountable factors, it is common that particle pH has variation when simulated with the ambient data. Although the pH predicted by the thermodynamic models are of uncertainty, it is widely believed that haze particles in China are moderately acidic and are more neutralized than those in the US, given that gaseous ammonia is still at a high level relative to particulate ammonium (Song et al., 2018).

Air pollution control in China has entered the second phase – further mitigation of the moderate haze pollution, which is characterized by high levels of nitrate and ammonium and a low level of sulfate (Liu et al., 2019; de Foy et al., 2016) due to the efficient SO2 emission control. Such a change in chemical compositions could significantly alter physicochemical properties of the atmospheric aerosols in China. This paper aims to investigate the variation in the particle acidity of PM2.5 as sulfur emission is well controlled and nitrogen oxide emission remains high in Beijing. First, the compositions of air pollutants, including inorganic components of PM2.5 during the winter of 2017 and 2018, were analyzed and compared with previous studies; then, based on observations, the response of particle acidity to the elevation of nitrate was studied by using the ISORROPIA II thermodynamic model; finally, possible changes in the future are discussed based on the sensitivity tests.

The observations were conducted at an urban site – the State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (39∘58′28′′ N, 116∘22′16′′ E) in Beijing. All the instruments were on the roof of a two-story building. The local emissions are mainly from vehicles, and the industrial emissions are greatly reduced since the major factories and power plants have been moved out of Beijing or phased out due to the emission control policy. Overall, this site represents the typical atmospheric environment in urban Beijing, and the data obtained here can be compared with those from previous studies in the city (Ji et al., 2018).

A continuous online measurement of atmospheric components was conducted with a time resolution of 1 h. Two TEOM™ continuous ambient PM monitors using a PM2.5 or PM10 cyclone inlet (Met One) were applied to obtain PM2.5 and PM10 mass concentrations (Tang et al., 2016). For trace gases (O3, NO2, and SO2), a series of gas monitors were used for the hourly measurement (models 49i, 42i, and 43i, respectively) (Tang et al., 2012). Meteorology data, including ambient temperature, RH, wind speed, wind direction, and total solar radiation, were measured with an automatic weather station (MILOS520, Vaisala Inc., Finland) located in the middle of the observation site yard. Visibility data of Beijing were downloaded from the open database (https://gis.ncdc.noaa.gov/maps/ncei/cdo/hourly, last access: 27 July 2018). Apart from these online monitors, a high-volume sampler (Tisch Environmental) with a PM2.5 inlet was used to collect PM2.5 samples on a day–night basis (daytime at 08:00–17:50 and nighttime at 18:00–07:50).

The inorganic water-soluble components of PM2.5 (SO42-, NO3-, Cl-, NH4+, Na+, Ca2+, Mg2+, and K+) and ammonia gas were measured with an online-IC system: IGAC (in situ gas and aerosol composition monitor, Fortelice International Co., Ltd.). The IGAC is comprised of two parts: a sampling unit and an analyzer unit (Young et al., 2016). A vertical wet annular denuder (WAD) is used to collect the gas-phase species prior to a scrub and impactor aerosol collector (SIC), which can efficiently collect particles into liquid samples. During the campaign, 1 mM H2O2 solution is used as the absorption liquid for the air samples. Under most atmospheric conditions, the absorption liquid can efficiently absorb the target atmospheric components (e.g., SO2). An ICS-5000+ ion chromatograph is used as the analyzer unit in this study. For anions, an AS18 column (2mm×250mm, Dionex™ IonPac™) is used while a CS-16 column (4mm×250mm, Dionex™ IonPac™) is chosen to analyze major cations, both running with the recommended eluent (solution of KOH for anion and methane sulfonic acid for cation). The performance of the IGAC system has been tested and improved over recent years, and studies of PM2.5 water-soluble ion observations have been conducted by using it (Young et al., 2016; Song et al., 2018; Liu et al., 2017). A better sensitivity due to the advanced suppression technology of the system greatly enhances its ability to measure trace ions, such as sodium and magnesium, which is important in studies of particle ion balance. For details of the comparison between IGAC and filter sampling results, please refer to the Supplement (Fig. S1).

So far, the effect of emission control on SNA compositions in Beijing's PM2.5 pollution and the response of particle pH have been illustrated, but it is important to make predictions with the current knowledge, providing scientific evidence for future better control strategy. In this section, sensitivity tests regarding hygroscopicity and particle-acidity change are conducted to help understand the possible changes of these properties in the future. ALWC is directly engaged in the calculation of particle pH and limited by several major parameters (RH, hydrophilic composition concentration, temperature, etc.). During the campaign, the ALWC predicted by ISORROPIA II varied between 0.8 and 154.2 µgm-3 with an average value of 6.4 µgm-3. As previously mentioned, the average value of RatioN-to-S is around 2 during the haze events in the winter of 2017 in Beijing. The average ALWC in the haze events increased to 24.4 µgm-3 accordingly. It has been reported that nitrate salts have a greater contribution to ALWC due to its lower deliquescence RH, and the elevated ALWC might have a strong impact on water-soluble gas partitioning such as that of glyoxal, leading to enhanced SOA production (Hodas et al., 2014; Xue et al., 2014). As a matter of fact, the increase in hygroscopicity related to nitrate-rich fine particles has been observed in Beijing (X. Wang et al., 2018). However, it is difficult to conclude that lower pH in nitrate-rich particles is caused by the dilution of H+ with higher ALWC with current data, since the higher nitrate fraction is usually observed with moderate RH in pollution episodes in the winter of 2017. Furthermore, the elevation of pH due to the RatioN-to-S increase is one unit of pH, which means the ALWC shall increase to 10 times the amount. This assumption is not supported by current data, since ALWC remains less varied as RatioN-to-S increases (Fig. S6). Similar to ALWC, the correlation between ambient temperature and RatioN-to-S is low, further proving that the increase in pH is not caused by change of thermodynamic state (Fig. S7).

The possible enhancement of hygroscopicity in nitrate-rich PM2.5 was investigated. Most single salts can only be deliquesced over a certain RH; thus the ALWC only exists when a certain RH is exceeded (Wexler and Seinfeld, 1991; Mauer and Taylor, 2010). In the real atmosphere, aerosols are usually a mixture of salts and organics, which might be easier to deliquesce. In addition, the deliquescence of NH4NO3 is unique, and it becomes more complicated in the system of NH4NO3/(NH4)2SO4. The NH4NO3 system would absorb water vapor even at an extreme low RH, e.g., down to 10 % (Willeke et al., 1980; ten Brink and Veefkind, 1995). Previous studies show that the system comprised of NH4NO3/(NH4)2SO4 has a higher deliquesce point when the sulfate content is higher (RatioN-to-S<1) but absorbs water even at low RH (∼20 %) when nitrate is dominant (RatioN-to-S>2) in PM2.5 inorganic ions (Wexler and Seinfeld, 1991; Ge et al., 1996, 1998). Inspired by these facts, we conducted a sensitivity test of ALWC to RH by using the observation dataset to study the effect of nitrate fraction elevation on ALWC changes (the RH value ranging from 20 % to 90 %, with 10 % as the interval). Concentrations of pollutants in the clean periods are relatively low and the data of the clean periods might be more influenced by the observation artifacts. Thus, only the data obtained in the transition and pollution conditions were analyzed here. The ALWC changes are defined as Eq. (1). 1Fractionchange=ALWC(RH + 10 %)/ALWCRH Then, we choose the data with RatioN-to-S above 3 and RatioN-to-S below 1. These values were both mentioned in previous lab studies (Ge et al., 1998) and are also typical values of nitrate-rich or sulfate-rich conditions in field observations. As shown in Fig. 10, PM2.5 with higher RatioN-to-S adsorbs more water than a lower nitrate fraction as the RH increases, which is more significant under lower RH (<50 %) conditions compared with that under higher RH (50 %–70 %) conditions. As the RH is usually lower (30 %–50) at the beginning stage of PM2.5 pollution development in Beijing, such a significant increase in hygroscopicity of nitrate-rich particles can greatly promote the haze formation under relatively dry conditions by enhancing the gas-to-particle partitioning of water-soluble compounds and the aqueous-phase formation of secondary aerosols, e.g., ammonia partitioning and nitrate formation through partitioning or hydrolysis of N2O5 (Badger et al., 2006; Bertram and Thornton 2009; Sun et al., 2018; Hodas et al., 2014; Shi et al., 2019).

The response of particle pH to ammonia and sulfate changes has been studied previously (Weber et al., 2016; Guo et al., 2017; Murphy et al., 2017). Here we further analyze the particle pH under the elevated nitrate concentration with the increasing ammonia in the atmosphere, which is the possible situation for most Chinese cities in the coming years. Two kinds of pH sensitivity tests are conducted: one with fixed nitrate but varying sulfate and ammonia and the other with fixed sulfate input but varying nitrate and ammonia (Fig. 11). In the test, crustal ions were all set as 0, while fixed chloride, sulfate, and nitrate concentrations were set as the average data on pollution (see Table 2). Compared with previous studies (Guo et al., 2017; Song et al., 2018), the RH was set as 58 % and the temperature was set as 273.15 K. Despite system errors due to the instability of the model at the extreme high-anion and low-NHx conditions (Song et al., 2018), the pH continuously changes as the free variable changes. The significant sharp edge of pH values in both plots defines the ion balance condition. We selected the observation data obtained during the pollution episodes within the RH ranging from 50 % to 70 % to compare with the results of both sensitivity tests. As shown in Fig. 11, apart from some data points (those with lower nitrate concentration but very high NHx concentration), observation data (triangles) are quite well merged into the results of sensitivity tests, and the pH values are generally higher than the test results due to the lack of crustal ion input in the sensitivity simulation. Therefore, the result of sensitivity tests can represent the pH change of the real atmosphere environment in Beijing well.

Future changes in particle pH can be found with the sensitivity test results. Cutting down the sulfate concentration without reducing atmospheric ammonia (horizontally moving from right to left in Fig. 10, left part) can lead to a significant increase in particle pH (up to 5). As can be seen from the right part of Fig. 10, the elevation of particle pH might be enhanced with the help of more nitrate in PM2.5. The effect of nitrate on particle pH greatly relies on the NHx concentration in the atmosphere. As the ammonia in the atmosphere over north China might still be increasing (Liu et al., 2018), and the sulfur content in the atmosphere might not be greatly reduced in the future, the particle pH shall increase in the path along the ion balance edge, which also implies a synergetic effect of increased nitrate and ammonia.

These results (lower acidity, higher hygroscopicity) provide insights into the effects of an elevated nitrate content on the physiochemical properties of particles. First, heterogeneous reactions that do not need high acidity might greatly contribute to the airborne particle chemistry, such as NO2-induced oxidation of the SO2 mechanism (Cheng et al. 2016; Wang et al., 2016). Reactions which rely greatly on acidified particles might contribute less, such as acid-catalyzed SOA formation from VOCs (Jang et al., 2002; Surratt et al., 2010). Second, the uptake processes of gaseous compounds onto particles (carbonyl acids, for example) might be enhanced, and the uptake of alkaline compounds could also be enhanced via the ALWC elevation. Third, optical properties of particles will greatly vary. On the one hand, higher ALWC can increase the light-scattering effect (Titos et al., 2014), while on the other hand the light absorption by BrC would be enhanced at higher pH (Phillips et al., 2017). All these facts might result in difficulties controlling moderate haze in Beijing, which usually occurs with lower RH and higher nitrate content as shown in this study. It is strongly suggested that the control strategy should be created accordingly based on thorough and scientific evaluation of both NOx and ammonia.

Due to strict emission controls, PM2.5 in Beijing during the winter of 2017 greatly decreased to a low level (39.7 µgm-3 for average concentration), but moderate haze episodes still frequently occurred in the city. With the observation and historical data, we found that the SO2 concentration decreased significantly while the NO2 concentration far exceeded that of SO2 and kept increasing in Beijing during winter. In response to the emission control, the nitrate concentration exceeded the concentration of sulfate significantly and thus became the dominant SIA component in fine particles. The molar ratio of nitrate to sulfate kept increasing over the years and rose to 2.7 during PM2.5 pollution episodes in the winter of 2017. The ammonium-to-sulfate ratio has always been above 1.5 in Beijing and has exceeded 3.0 since 2011. Sufficient ammonia provided strong atmospheric neutralization and weakened the particle acidity in Beijing, but the increased nitrate fraction was found to be causing the particle pH elevation. During the campaign, the pH of PM2.5 increased from 4.4 to 5.4 as the molar ratio of nitrate to sulfate increased from 1 to 5, which is firstly due to the lower amount of sulfate, which suppressed the formation of H+, and secondly due to the ammonia neutralization. Sensitivity tests of particle hygroscopicity and acidity were conducted to investigate the possible changes in physiochemical properties if ammonia and nitrate are not well controlled in China in the future. The results showed that the nitrate-rich particles can absorb more water than particles with higher sulfate fractions under moderate humid conditions (RH <60 %), and the particle pH increases rapidly due to the synergetic effect of ammonia and nitrate, which will very likely occur in China in the upcoming years, because both of these pollutants are not well controlled yet. The changes in particle pH and hygroscopicity will further enhance the uptake of gaseous compounds, promote chemical reactions which favor lower acidity, and also affect the optical properties of airborne particles in China. Therefore, the processes and properties of haze particles during nitrate-dominated periods in the country need to be thoroughly investigated with more consideration of highly hygroscopic and neutralized particles.