Variation of Urban Atmospheric Ammonia Pollution and its Relation with PM2.5 Chemical Property in Winter of Beijing, China

To understand the air pollution problem in megacities such as Beijing, field measurement investigating the variation of NH3 and its association with PM2.5 chemical property was conducted from 25 November to 24 December 2013. The results indicated that the daily concentration of wintertime NH3 tended to be high on the days with relatively high temperatures and low wind speeds. Affected by the synoptic condition, NH3 concentration showed a bimodal diurnal variation pattern, which tended to peak at around 09:00 and 22:00 of the day. As the sole precursor for NH4, NH3 exerted a significant impact on the ion chemistry of PM2.5 through enhancing the nighttime NH4Cl formation and promoting both homogeneous and heterogeneous formation of NO3. During heavy pollution episodes with PM2.5 concentrations over 200 μg m, the NH3 levels and NH4/NH3 ratios grew simultaneously with the increase of PM2.5 levels, indicating that NH3 is one of the key reasons for heavy pollution events. Revealed by the features of measured ionic species in PM2.5, in conjunction with the acidity analysis using thermodynamic model, our results suggested that NH3 was frequently sufficient in wintertime atmosphere of urban Beijing and the fine particulates were neutralized nearly fully by NH3.


INTRODUCTION
Beijing, the capital city of China, has been one of the most polluted cities in Asia in recent years.One of the major air pollutants is PM 2.5 (fine particles with aerodynamic diameters less than or equal to 2.5 µm), which remains a nationwide problem despite considerable efforts for its removal (Lu et al., 2015).Accounting for a large fraction of PM 2.5 mass, sulfate (SO 4 2-), nitrate (NO 3 -), and ammonium (NH 4 + ) are the main secondary inorganic ions in fine particulate matters (He et al., 2001;Wang et al., 2005;Li et al., 2013;Zhang et al., 2013a;Chen et al., 2014).The ambient precursor gases, i.e., SO 2 , NO x , and NH 3 , are responsible for the formation of secondary inorganic components in PM 2.5 though the conversion extent may largely depend on the condition of atmospheric environment.Among the three major precursor gases, SO 2 and NO 2 are routinely measured by the National Air Quality Monitoring Networks in China whereas NH 3 is not.As a dominant base species in ambient air, NH 3 plays a significant role in atmospheric chemistry.Besides its close relationship with particulate formation in air, however, NH 3 also has distinct impacts on the soil acidification and eutrophication of ecosystems (Behera et al., 2013a).
The needs to reach a better understanding of atmospheric NH 3 in recent years have increased, especially in urban areas where air pollutants involve a more complex mixture of harmful gases and particles of varying origins and chemical compositions, including emissions from transportation, commercial, industrial, domestic and fugitive activities and secondary formation from atmospheric chemical processes.In some cities around the world, measurements of ambient NH 3 have been reported, for example, in Rome (Perrino et al., 2002), New York City (Li et al., 2006), Manchester (Whitehead et al., 2007), Barcelona (Pandolfi et al., 2012) and Seoul (Phan et al., 2013).In China, continuous measurements of NH 3 spanned over a year have been reported in Beijing (Meng et al., 2011) and Xi'an (Cao et al., 2009), which focused on characterizing the levels and variations of NH 3 .However, what concerned people more was the impact of atmospheric NH 3 on particulate pollution.Ye et al. (2011) studied the haze formation in Shanghai and found that the NH 3 played a vital role in the enhancement of particulate sulfate and nitrate during the haze episode.Gong et al. (2013) found that elevated NH 3 levels were synchronous with enhancements in NH 4 + and SO 4 2-around midday in Houston during summertime and suggested NH 3 likely influenced both mass and number concentrations of particles.Toro et al. (2014) concluded that the atmosphere of Santiago city can be considered to be NH 3 -rich since abundant NH 3 was present there to neutralize the acid components to form fine particulate ammonium salts.
However, although Beijing is repeatedly shrouded by haze episodes, studies in literature barely reported temporal variation of NH 3 and its synchronous relation with PM 2.5 chemical property in the city.Particularly, among the four seasons, Beijing suffered from much more severe and frequent haze episodes in the winter season (Zhao et al., 2013;Jiang et al., 2015).The secondary inorganic aerosol formation could be a major reason for the wintertime PM 2.5 pollution and the sum of SO 4 2-, NO 3 -, and NH 4 + can occupy 50.3-53.2% of the PM 2.5 mass in haze episodes (Zheng et al., 2015;Liu et al., 2016).Under this background, special attention should be paid to discern the mechanism of the secondary inorganic aerosol formation in winter and it is worth investigating the behavior of NH 3 in this season to help craft effective management strategies to mitigate the air pollution problem.In the present study, we carried out a continuous 30-day field measurement at an urban site of Beijing in the winter of 2013, simultaneously collecting data of NH 3 , NO x , and chemical species of PM 2.5 .The aim of this study was to characterize the wintertime NH 3 variation and elucidate the role of NH 3 played in the ion chemistry of PM 2.5 in the real atmosphere.The results represent an important contribution to understand the air pollution in Beijing from the perspective of NH 3 , an important alkaline and active gas in ambient air.

Field Measurement
The measurement site (39°54'N, 116°15'E) is located on the campus of University of Chinese Academy of Sciences situated in Shijingshan District of Beijing.It is in the western urban area of Beijing and between the Fourth Ring Road and the Fifth Ring Road (Fig. 1).This site can be categorized as a typical urban site in Beijing.
From 25 November to 24 December 2013, a model 17i ammonia analyzer (Thermo Fisher Scientific Inc., USA) was employed to monitor the NH 3 and NO x (NO + NO 2 ) in ambient air with a time resolution of 1 min, which can show a rapid response to the ambient level of NH 3 .The ammonia analyzer was deployed on the ninth floor of one building with a height of about 25 m above the ground.It was continuously running except for occasional calibration and maintenance.The stated manufacturer precision of the ammonia analyzer is ± 0.3 µg m -3 .Model 17i ammonia analyzer is a chemiluminescent gas analyzer and it uses the light producing reaction of nitric oxide with ozone as its basic principle.
In parallel with NH 3 monitoring, PM 2.5 samples were manually collected by a model Partisol 2300 speciation sampler (Thermo Fisher Scientific Inc., USA) using denuder/backup-filter sampling method, which helped to minimize the sampling artifacts of the semivolatile species (e.g., NH 4 NO 3 and NH 4 Cl) in PM 2.5 (Pathak et al., 2004;Yu et al., 2006;Huang et al., 2011).The PM 2.5 sampler was installed on the rooftop of another building (adjacent to the building for NH 3 monitoring) with a height of about 16 m above the ground and operated at a constant flow rate of 10 L min -1 .Samples were collected daytime and nighttime separately with each sampling period last 12 h, from 8:00 to 20:00, or from 20:00 to 08:00 of the next day.Two of the four channels were used.In one channel, three filters for different purposes were placed, that is, the first Teflon filter (Whatman, 47 mm diameter) was set up to collect PM 2.5 while the second nylon filter (Pall, 47 mm diameter) and the third quartz filter (Pallflex) were designed to absorb acid gases and NH 3 respectively, which volatilized from particles collected on the front Teflon filter.To prevent related gases in the air from interfering the backup filters' absorption of evaporated gases, two glass honeycomb denuders were placed in front of the three filters to eliminate them.Of the two denuders, one was coated with 20 g L -1 Na 2 CO 3 solution and the other was coated with 20 g L -1 citric acid solution.The solvent of the Na 2 CO 3 solution was a 1:1 volume mixture of methanol and deionized water (18 MΩ cm) while that of the citric acid solution was methanol only.Coated denuders were dried following coating by using a high purity N 2 system before use.In the other channel that used, a single quartz filter was put to collect PM 2.5 for measuring carbonaceous components.It should be noted that before sampling, all the quartz filters were preheated at 550°C for 5.5 h, and then the backup ones were impregnated with the 20 g L -1 citric acid solution and dried in a high purity N 2 environment.Totally 59 sets of filter samples, i.e., a number of 236, were collected throughout the sampling period.No samples were collected during the night of 12 December due to a power outage.In addition, each sample was sealed and refrigerated immediately after the end of its sampling.The artifacts during the sampling and analysis were estimated by field blank filters.
Meteorological parameters including temperature, relative humidity, and pressure were simultaneously recorded by the sampler.Daily average wind speed data were obtained from China Meteorological Data Sharing Service System (http://cdc.nmic.cn/home.do).Wind speed and wind direction data at 02:00, 08:00, 14:00 and 20:00 of the day were downloaded from Weather Underground (http://www.wunderground.com/).Planetary boundary layer (PBL) height data in 3 h time resolution were obtained from the ARL (Air Resources Laboratory) at the NOAA website (http://www.arl.noaa.gov/index.php).The validity of NOAA PBL data has been verified through comparison with vertical lidar observations (Ungureanu et al., 2010).There was no precipitation during our observation.

Sample Analyses and Quality Assurance
Mass concentrations of PM 2.5 were determined by weighing Teflon filters before and after sampling using an analytical balance (Mettler Toledo AX105DR).Prior to weighing, filters were kept in a conditioned room for at least 24 h at a relative humidity of 40 ± 3% and a temperature of 20 ± 1°C.Five cations (Na + , NH 4 + , K + , Mg 2+ , and Ca 2+ ) and three anions (SO 4 2-, NO 3 -, and Cl -) were analyzed in aqueous extracts of the filters by ion chromatographs (ICS-1000 for cation and ICS-2000 for anion, Dionex, USA).To extract the water-soluble ions from the filters, each sample was extracted ultrasonically by deionized water (18 MΩ cm) and filtered through microporous membranes with pore size of 0.45 µm.The eight ionic species were identified and quantified by interpolation on the constructed calibration curves that all had a linear correlation coefficient R 2 above 0.996.More details about the ionic analysis have been described elsewhere (Wang et al., 2006).It should be noted that the reported particulate NH 4 + concentration in this study was the sum of the measurements on Teflon filter and citric acid coated quartz filter while the concentrations for NO 3 -and Cl -were both the sum of the measurements on Teflon filter and nylon filter (Huang et al., 2011).On this basis the average aerosolmode losses of NH 4 + , NO 3 -, and Cl -were 6.5%, 8.1%, and 14.9%, respectively, comparable to the corresponding values of 6%, 7%, and 22% reported by Zhang et al. (2013b).
The procedures to extract sampled gases from the honeycomb denuders strictly followed the manual of Partisol 2300 speciation sampler.Although the concentrations of NH 3 determined from denuder using ion chromatography were comparable with those measured by ammonia analyzer on some days, the efficiency of the denuder in the present study remained much uncertain due to the required fussy operations (He et al., 2001).Norman et al. (2009) and Meng et al. (2011) compared different ammonia analysis methods and found accuracy of the online method is acceptable (correlation coefficient ranged from 0.77-0.94).Moreover, the chemiluminescence method has been used in several studies to measure ambient NH 3 and has exhibited satisfactory performances (Behera et al., 2010;Meng et al., 2011;Pandolfi et al., 2012).To obtain a finer temporal view of NH 3 variation, the NH 3 data from the ammonia analyzer was used in this study.
A punch of 0.5 cm 2 from each quartz filter (those without a coating of citric acid solution) was analyzed for OC and EC using a thermal/optical carbon analyzer (DRI 2001, USA), following the IMPROVE_A protocol.OC is operationally defined as OC1 + OC2 + OC3 + OC4 + OP and EC is defined as EC1 + EC2 + EC3 -OP, where OP is the pyrolyzed OC.In addition, the concentration of organic matter (OM) is subsequently estimated by multiplying OC with a conversion factor of 1.55 (Cheng et al., 2014).Quality assurance and quality control were conducted according to Li et al. (2015).

Calculation of PM 2.5 Acidity
The in-situ acidity of PM 2.5 was estimated by a chemical thermodynamic model ISORROPIA-II that treats the K + -Ca 2+ -Mg 2+ -NH 4 (Fountoukis and Nenes, 2007).To obtain the best available predictions of aerosol pH, ISORROPIA-II was run in the forward mode with metastable aerosol state (Hennigan et al., 2015).Concentrations of the measured inorganic species were input into the model as total (gas + aerosol) concentrations, along with relative humidity and temperature data.The gaseous fraction of the input NH 4 + was determined by the 17i ammonia analyzer measurement, whereas those of the input NO 3 -and Cl -were determined by the aqueous extracts of denuders using ion chromatography.Fig. 2 showed the comparison of predicted NH 3 (output of the model) with the measured NH 3 .They were in good agreement except for a few days, with a normalized mean bias (NMB) of -23.5% and a correlation coefficient (R) of 0.59 for all the data.This result provided confidence in ISORROPIA-II calculations of particle pH (Guo et al., 2015).

NH 3 Concentration and Its Temporal Variation
The temporal variations of daily averaged NH 3 concentrations and meteorological parameters during the measurement period were shown in Fig. 2.During this period, the daily averaged NH 3 concentrations varied from 7.6 to 38.1 µg m -3 , with a mean value of 18.2 µg m -3 and a median value of 16.2 µg m -3 .The maximum value was about 5 times higher than the minimum one, which showed a wide variation range of NH 3 concentration in urban Beijing.Daily NH 3 exhibited a distinct and significant temporal variation with higher concentrations on the days with higher temperatures and lower wind speeds.High temperatures will favor NH 3 volatilization from sources present in the study area and low wind speeds will lead to accumulation of air pollutants.NH 3 concentration observed in this study was higher than that of 0.20-14.08µg m -3 in the winter of 2007 reported by Ianniello et al. (2010) at the Peking University site (urban site) in Beijing, and was comparable to the annual mean value of 17.8 µg m -3 in 2009 obtained at another urban site in Beijing reported by Meng et al. (2011).It should be noted that due to high temperatures and stable atmospheric conditions, several days with relatively high NH 3 levels were recorded in our observation, consequently leading to a relatively large mean NH 3 concentration.The NH 3 level in Beijing was lower compared to the measurements in some other cities around the world; for example, in Kaohsiung, Taiwan (31.5 ± 16.1 µg m -3 , winter 2006, Tsai et al., 2014); Kanpur, India (26.4 ± 5.8 µg m -3 , winter 2007, Behera and Sharma, 2010); and Lahore, Pakistan (50.1 ± 16.9 µg m -3 , winter 2005, Biswas et al., 2008).However, in Beijing, the NH 3 level was still higher than those observed in some other cities of China including Shanghai (6.6 µg m -3 , autumn 2012, Shi et al., 2014), Xi'an (12.88 µg m -3 , April 2006-April 2007, Cao et al., 2009), and Guangzhou (7.3 µg m -3 , autumn 2004, Hu et al., 2008).The relatively high NH 3 level observed during the measurement period in conjunction with the minor molar mass of NH 3 (17 g mol -1 ), signified the important role that NH 3 could play in atmospheric chemistry in urban Beijing.
Fig. 3 illustrated the diurnal variations of hourly NH 3 concentrations measured during the entire observation campaign.On average, a bimodal diurnal variation pattern of NH 3 concentration was found in Beijing urban area in winter, consistent with that observed at another urban site of Beijing in the winter of 2009 (Meng et al., 2011).From Fig. 3, it can be seen that on most of the days, NH 3 concentration tended to peak at around 09:00 in the morning and at around 22:00 in the midnight of the day.Troughs of NH 3 concentration were mostly seen at around 06:00 in the early morning and at around 16:00 in the afternoon.However, it seemed that no remarkable differences occurred between NH 3 diurnal cycles during the weekdays and weekends.The weekday versus weekend differences in NH 3 profiles indicated that vehicular emission had a relatively small impact on ambient NH 3 at the measurement site, which was consistent with the ammonia emission inventories of Beijing.According to Huang et al. (2012) and Zhou et al. (2015), more than 60% of total ammonia emission in Beijing comes from livestock and farm-land.Other sources, including human excrement, waste disposal, biomass burning, chemical industry and traffic, totally contributed 14.9-35.5% to the total budget with vehicular source accounting for only about 5%.Expectedly, on a few days the scavenging or accumulating process affected the diurnal variation pattern of NH 3 concentrations.
To gain a further insight into how this NH 3 diurnal variation was formed, the relationship between NH 3 and key meteorological parameters was examined.Table 1 showed the summary of meteorological parameters and their Pearson correlations with NH 3 .Among the three parameters, temperature was found to have a significant positive correlation with NH 3 while the other two, wind speed and PBL height, had significant negative correlations with NH 3 .This result corresponded with the fact that high temperature facilitates NH 3 release from its sources and both low wind speed and low PBL height cause intensive atmospheric stability, which limits the dispersion of air pollutants.On average, the daytime temperature was about 3.50°C higher than that at nighttime.Similarly, both wind speed and PBL height diminished from daytime to nighttime, with from 2.53 to 1.77 m s -1 and from 0.50 to 0.22 km for wind speed and PBL height, respectively.Hence, the morning NH 3 peak could be due to emissions from dew evaporation with increasing temperature around sunrise; enhanced release from biological sources in the city such as human excreta, waste disposal, and vegetation; and also regional diffusion from adjacent suburbs (Huang et al., 2012).In late morning and afternoon, the development of PBL height, i.e., dilution effect, as well as high wind speed, would be the reason for the decrease trend of NH 3 during that time period.The night NH 3 peak was likely the result of accumulation due to lower PBL heights and more frequent temperature inversions at winter nights.However, decreasing of NH 3 level from midnight to early morning might be related to the relatively small amount of nighttime emissions under low temperature and the enhanced participation of NH 3 in particulate formation.All above results suggested that the diurnal variation of NH 3 was notably affected by the synoptic condition or more specifically by the combined effects of temperature, wind speed, and PBL height.

Relationship between NH 3 and Ammonium Salts in PM 2.5 Overall Relationship between NH 3 and Fine Particulate Ammonium Salts
In air, NH 3 is the precursor gas of NH 4 + in particles.The gas-to-particle conversion processes in ambient atmosphere may produce inorganic ammonium salts of ammonium bisulfate (NH 4 HSO 4 ), ammonium sulfate ((NH 4 ) 2 SO 4 ), ammonium nitrate (NH 4 NO 3 ), and ammonium chloride (NH 4 Cl) (Tsai et al., 2014).It is known that NH 4 HSO 4 or (NH 4 ) 2 SO 4 is preferentially formed among the ammoniumassociated compounds because the affinity of sulfuric acid for NH 3 is much larger than that of nitric acid and hydrochloric acid for NH 3 (Pandolfi et al., 2012;Behera et al., 2013b).The excess available NH 3 may react with nitric acid and hydrochloric acid to form NH 4 NO 3 and NH 4 Cl; however, the reaction rate constant of NH 3 with nitric acid (1.59 × 10 -4 m 3 µmol -1 s -1 ) is much higher than that with hydrochloric acid (5.16 × 10 -5 m 3 µmol -1 s -1 ) (Behera and Sharma, 2012).
The overall relationship between molar concentrations of NH 3 and the corresponding PM 2.5 -bound NH 4 + during the whole field measurement period was shown in Fig. 4(a).In general, they had a good linear relationship with a correlation coefficient R 2 of 0.61 and a slope of 0.48, suggesting the important precursor role of NH 3 played in the NH 4 + formation in PM 2.5 .However, the formation of NH 4 + does not solely depend on NH 3 concentration.It also depends on several chemical and meteorological factors, such as the concentrations of atmospheric acidic gases, characteristics of pre-existing aerosols, air temperature, and humidity.Meteorological analysis showed that NH 3 to NH 4 + conversion was favored under the conditions of low temperature and high relative humidity (see Fig. 5).These findings implied high NH 3 level under the winter synoptic condition might be one of the important factors that facilitated fine particulate ammonium salt formation during haze episode in urban Beijing.
To further assess the availability of NH 3 for neutralizing the atmospheric acidic substances, the associations between NH 4 + and anions in PM 2.5 were investigated and were illustrated in Fig. 4(b).It can be seen that the average molar ratio of NH 4 + to 2SO 4 2-(each mole of SO 4 2-removes 2 moles of NH 4 + ) was 2.08, indicating the complete neutralization of sulfuric acid and a predominance of (NH 4 ) 2 SO 4 in sulfate salts during the winter season.On the whole, NH 4 + appeared to almost completely neutralize SO 4 2-and NO 3 -because the average molar ratio of NH 4 + to (2SO 4 2-+ NO 3 -) was 1.05.In terms of NH 4 + versus 2SO 4 2-+ NO 3 -, about 67.2% of the data points were above the 1:1 line, signifying a possible frequent existence of NH 4 Cl in PM 2.5 .A study carried out by Behera et al. (2013b) suggested that NH 4 Cl formation was not favored under NH 3 -poor conditions.This meant the winter atmosphere of urban Beijing was frequently under sufficient NH 3 condition and the process of neutralization could possibly form NH 4 Cl besides (NH 4 ) 2 SO 4 and NH 4 NO 3 .The strongest correlation coefficient found between NH 4 + and 2SO 4 2-+ NO 3 -+ Cl -among the three cases in Fig. 4(b) also confirmed this.

Relationship between NH 3 and NH 4 + during Typical PM 2.5 Pollution Episodes
In the present study, significantly high levels of PM 2.5 with daily concentration above 200 µg m -3 were observed on 7, 22 and 24 December.Daily PM 2.5 concentrations on these three days were recorded as 257.5, 222.6 and 269.5 µg m -3 , respectively, about 6-7 times of the Chinese National Ambient Air Quality Standard (35 µg m -3 for daily PM 2.5 , GB 3095-2012).To specifically explore the behavior of ambient NH 3 on the days under severe PM 2.5 pollution, two episodes were studied: episode 1 from 5 to 9 December and episode 2 from 20 to 24 December.Fig. 6 presented the evolution of PM 2.5 chemical composition and gaseous NH 3 and NO 2 during episode 1 and episode 2, together with the mass ratios of NH 4 + /NH 3 and NO 3 -/NO 2 .Obviously, in episode 1, before the peak on 7 December, increasing PM 2.5 concentrations were observed, followed by a sharp decrease on 8 December.In episode 2, PM 2.5 concentration increased from 20 December and reached the first peak on 22 December, then after a moderate decrease on 23 December it reached the second peak on 24 December.As major PM 2.5 components, OM, EC, SO 4 2-, NO 3 -, and NH 4 + showed very similar fluctuations with PM 2.5 .From 5 December to 7 December, OM, EC, SO 4 2-, NO 3 -, and NH 4 + increased 3.0, 2.1, 43.2, 22.1, and 6.3 folds, respectively; during 20-24 December, OM, EC, SO 4 2-, NO 3 -, and NH 4 + increased 6.3, 4.0, 17.2, 21.1, and 6.1 folds, respectively.Such increase was the major cause of PM 2.5 increase.With relatively low daily average wind speeds (1.1-1.4 m s -1 ) and PBL heights (0.11-0.15 km) on 7, 22, and 24 December (Fig. 2), the stable synoptic conditions which favored accumulation of emitted pollutants and rapid formation of secondary species were essential to the formation of the severe PM 2.5 pollution (Zheng et al., 2015).The high daily average relative humidity (58.3-81.0%) on these three days also favored secondary inorganic aerosol formation through hygroscopic growth of ambient particles and heterogeneous reactions of gaseous precursors (Ye et al., 2011).Concentrations of K + and Cl -, influenced by anthropogenic activities such as biomass burning (Cheng et al., 2014) and coal combustion (Evans et al., 2011), also showed obvious increases during the two episodes.However, the crustal species Ca 2+ , Mg 2+ , and Na + in PM 2.5 did not tend to build up on heavily polluted days.
It should be noted that during both the two episodes, when PM 2.5 concentrations gradually elevated in the first three days, the NH 3 concentrations also showed an upward trend at the same time.Peaks of NH 3 and PM 2.5 levels were coincided on 7, 22, and 24 December, with NH 3 concentration reaching as high as 38.1 µg m -3 on 7 December.Meanwhile,  the NH 4 + /NH 3 ratios on the most polluted days in each episode (i.e., on 7 and 24 December, respectively) were 0.45 and 0.70, respectively, about 5 and 4 times larger than those on most clean days in each episode (i.e., on 9 and 20 December, respectively).With the dramatic increases in NH 3 level and NH 4 + /NH 3 ratio, the result pointed out that via neutralization process NH 3 was largely responsible for the enhanced formation of secondary inorganic species in PM 2.5 on heavily polluted days in urban Beijing.High levels of precursor gases (NO 2 and NH 3 ) and the stagnant atmosphere provided favorable conditions for the gas-toparticle conversion processes, which was also supported by high NO 3 -/NO 2 ratios on heavily polluted days.Since NH 4 + was mostly associated with SO 4 2-and NO 3 -in PM 2.5 , the above results suggested that enhanced fine particulate ammonium salt formation due to available NH 3 could be an important cause of PM 2.5 pollution in the urban atmosphere of Beijing.

Effects of NH 3 on Chloride and Nitrate Ions in PM 2.5 Nighttime Enhancement of NH 4 Cl Formation
During the measurement period, we found that the concentration of Cl -in PM 2.5 at nighttime was averagely 3.39 ± 3.88 µg m -3 higher than that at daytime.By plotting the scatter plots of daytime and nighttime molar concentrations of NH 4 + versus (2SO 4 2-+ NO 3 -), it can be seen that nocturnal NH 4 + concentrations were more frequently higher than the amount equal to (2SO 4 2-+ NO 3 -) when compared with those in daytime, and this result was graphically highlighted in Fig. 7. Also note that in contrast with the data points at daytime, the nighttime ones were more concentrated in the top-right region of the figure that represented higher ion levels.The excess of NH 4 + , i.e., the part exceeding the amount equal to (2SO 4 2-+ NO 3 -), was most likely associated with Cl -and occurred in the form of NH 4 Cl although Cl - may partly be associated with crustal species (e.g., with Na + to form NaCl).During the winter season, the reversible reaction between NH 3 and hydrochloric acid in the forward direction was driven by the cold synoptic condition, especially during the night (Fig. 5).However, it seemed that the high NH 3 availability played a very important role in the formation of NH 4 Cl because particulate SO 4 2-concentration under irreversible reaction increased 1.80 ± 4.83 µg m -3 from daytime to nighttime while that of NO 3 -under reversible reaction increased 0.51 ± 5.00 µg m -3 , with excess of NH 3 supporting the form of NH 4 Cl.As a result, elevated Cl - concentrations with an average nighttime to daytime ratio of 1.8 in fine particles during the night were observed.

Promotion of Nitrate Ion Formation
The ratio of NO 3 -to SO 4 2-exhibited an increasing trend in Beijing in recent years (He et al., 2012;Chen et al., 2014), consistent with the effective control of SO 2 emissions (Zhao et al., 2013) and the rapid increase of the number of motor vehicles.As a major inorganic species in PM 2.5 , particulate NO 3 -can be primarily formed through the homogeneous reaction between gaseous HNO 3 and NH 3 and the heterogeneous reaction between NH 3 and the product from the hydrolysis of N 2 O 5 (Pathak et al., 2011;Shon et al., 2013).Generally, during daytime, the homogeneous formation is predominant, whereas during nighttime, heterogeneous formation occurring through the hydrolysis of N 2 O 5 on the surface of the pre-existing moist aerosols under relatively high humidity is predominant (Ianniello et al., 2011).
As SO 4 2-competes with NO 3 -for NH 4 + during its formation, the relationship between molar ratios of NH 4 + /SO 4 2and NO 3 -/SO 4 2-is used to investigate the pathway of NO 3 formation (Huang et al., 2011;He et al., 2012).Fig. 8(a) depicted NO 3 -/SO 4 2-against NH 4 + /SO 4 2-at different time of the day.At daytime, both the slope and correlation coefficient of the trend line were found to be larger than those at nighttime, indicating a more significant and stable homogeneous formation of particulate NO 3 -in PM 2.5 .In order to further inspect the potential influence of NH 3 level on particulate NO 3 -formation, scatterplots of NH 4 + /SO 4 2and NO 3 -/SO 4 2-under different NH 3 concentrations were also plotted, as shown in Fig. 8(b).Obviously, the ratios of NO 3 -/SO 4 2-tended to be high under relatively high NH 3 concentrations at most of fixed small NH 4 + /SO 4 2-ratio ranges, regardless whether the time was in daytime or nighttime.This finding indicates that sufficient NH 3 in air significantly promotes both the homogeneous and heterogeneous formation of nitrate in PM 2.5 .The most possible explanation would be that the products from the photochemical reactions of NO x and the hydrolysis of N 2 O 5 were depleted via the neutralization of NH 3 to form NH 4 NO 3 , and the replenishment of them

Impact of NH 3 Levels on PM 2.5 Acidity
While NH 3 can influence aerosol phase inorganic ions in PM 2.5 , it may also exert an impact on aqueous phase hydronium ion (H + ) in aerosols.The concentration of H + in aqueous aerosols, or pH, is an important aerosol property, which is mainly determined by the balance of acidic ionic components with basic ones.During our measurement period, the atmosphere of Beijing was frequently abundant with NH 3 .As a result, about one third of the PM 2.5 samples were found alkaline with pH values fluctuated between 7 and 8.The remainder of the PM 2.5 samples (n = 42) were acidic and had pH values ranging from 3.6 to 6.8 with an average of 5.5.
To assess the influence of NH 3 level on acidity of PM 2.5 , we varied the initial NH 3 input from 25% to 200% of the measured value while keeping the other inputs constant using the 42 acidic PM 2.5 sample data.Changes in the pH of PM 2.5 as a function of the NH 3 levels were shown in Fig. 9.When NH 3 concentrations increased from 25% to 100% of the measured values, a sharp increase in the pH values was observed.However, any further increase in NH 3 concentrations above the measured values (100% case) resulted in only a moderate increase in the pH values, even in the situation doubling the NH 3 concentrations.It is therefore clear that a higher NH 3 level, which will neutralize the acidic components in PM 2.5 , is conducive to a higher pH value.In the present study, on average, a ±25% perturbation in NH 3 level could lead to a 0.14 unit pH increase for 25% perturbation above the measured value (125% case) and a 0.23 unit pH decrease for a perturbation below the measured value (75% case).The relatively flat increasing tendency of pH with increasing NH 3 level perturbed above our measured NH 3 value reflected and supported the conclusion that sufficient NH 3 was frequently present in wintertime atmosphere of urban Beijing and the fine particulates were almost fully neutralized by NH 3 .

CONCLUSIONS
The present study dedicated to characterize the NH 3 variation in winter season and describe its association with PM 2.5 chemical property in urban Beijing.The daily NH 3 concentration in urban Beijing was found to exhibit a distinct temporal variation with higher values on the days with higher temperatures and lower wind speeds.Affected by the synoptic condition, NH 3 concentration showed a typical bimodal diurnal variation pattern with the two peaks respectively tending to appear at around 09:00 and 22:00 of the day.Apart from acting as the sole precursor role of NH 4 + formation, NH 3 also exerted a significant impact on Cl -and NO 3 -formation in PM 2.5 via neutralization process.Elevated NH 3 levels and NH 4 + /NH 3 ratios occurred when PM 2.5 levels also increased considerably, which signified that NH 3 caused the aggravation of pollution level in ambient air.Revealed by the features of measured ionic species in PM 2.5 , combined with the acidity analysis using thermodynamic model ISORROPIA-II in the forward mode, our results suggested that at most time NH 3 was sufficient in air and the fine particulates were almost fully neutralized by NH 3 in winter of urban Beijing.Nevertheless, additional work is needed to discern the impact of atmospheric NH 3 on number concentrations of particles in different sizes in the future.
Strategic Priority Research Program of the Chinese Academy of Sciences (XDB05020300), and National Natural Science Foundation of China (21221004).

Fig. 1 .
Fig. 1.Location of the measurement site in Beijing, China.

Fig. 2 .
Fig. 2. Temporal variations of predicted and measured NH 3 concentrations and meteorological parameters during the study period.T: temperature; RH: relative humidity; PBL: planetary boundary layer.

Fig. 4 .
Fig. 4. Relationship between (a) gaseous NH 3 and fine particulate NH 4 + and (b) NH 4 + and anions in PM 2.5 during the measurement period.The dotted line was the 1:1 line.

Fig. 8 .
Fig. 8. Molar ratios of NO 3 − /SO 4 2− versus NH 4 + /SO 4 2− in PM 2.5 at conditions of (a) different time of the day and (b) different NH 3 concentrations in urban Beijing.

Fig. 9 .
Fig. 9. pH predictions using ISORROPIA-II model by perturbing measured NH 3 concentrations.The 100% case represents the predictions resulted from in-situ measurement data input.Mean and median values are represented by the square and line inside each box, respectively; the top and bottom of each box represent the 75th and 25th percentile values, respectively; and the top and bottom of each whisker represent the 95th and 5th percentile values, respectively.

Table 1 .
Summary of meteorological parameters and their Pearson correlations with NH 3 .