Characteristics and Sources of Water-Soluble Ions in PM 2 . 5 in the Sichuan Basin , China

To track the particulate pollution in Sichuan Basin, sample filters were collected in three urban sites. Characteristics of water-soluble inorganic ions (WSIIs) were explored and their sources were analyzed by principal component analysis (PCA). During 2012–2013, the PM2.5 concentrations were 86.7 ± 49.7 μg m−3 in Chengdu (CD), 78.6 ± 36.8 μg m−3 in Neijiang (NJ), and 71.7 ± 36.9 μg m−3 in Chongqing (CQ), respectively. WSIIs contributed about 50% to PM2.5, and 90% of them were secondary inorganic ions. NH4 and NO3 roughly followed the seasonal pattern of PM2.5 variations, whereas the highest levels of SO4 appeared in summer and autumn. PM2.5 samples were most acidic in autumn and winter, but were alkaline in spring. The aerosol acidity increased with the increasing level of anion equivalents. SO4 primarily existed in the form of (NH4)2SO4. Full neutralization of NH4 to NO3 was only observed in low levels of SO4 + NO3, and NO3 existed in various forms. SO4 and NO3 were formed mainly through homogeneous reactions, and there was the existence of heterogeneous reactions under high relative humidity. The main identified sources of WSIIs included coal combustion, biomass burning, and construction dust.


Introduction
Water-soluble inorganic ions (WSIIs) are a major part of fine particles (PM 2.5 , particulate matter with an aerodynamic diameter less than 2.5 µm).Of the various components, secondary inorganic ions (SII), including sulfate (SO 4 2− ), nitrate (NO 3 − ), and ammonium (NH 4 + ), are the predominant species and account for more than 90% of WSIIs [1].Nationwide, SII contribute about 25%-48% to PM 2.5 mass, and are attributable to about 60% of the visibility reduction in China [2].Moreover, they also play important roles in atmospheric acidification and climate change [3,4].Characteristics of the WSII pollution in many cities of China have been studied, and the formation of SII has always been a focus [5].Sulfate is primarily formed through homogeneous gas-phase oxidation of sulfur dioxide, while heterogeneous transformation processes, i.e., metal-catalyzed oxidation, H 2 O 2 /O 3 oxidation, and in-cloud process, are also reported [6] (p. 348), [7].Both homogeneous reaction via NO 2 oxidation by OH radical and O 3 , and the heterogeneous hydrolysis of N 2 O 5 on preexisting aerosols, are important pathways of nitric acid formation [8].
four-channel sampler at a flow rate of 16.7 L/min (model: TH-16A, Tianhong Instrument Co., Ltd., Wuhan).The sampling process was carried out by trained staff of a local monitoring station, and the flow rate of the sampler was checked monthly using a bubble flow meter (Gilian Gilibrator 2, Sensidyne, US) to ensure collection efficiency.The sampled filters were stored in refrigerator at −18 °C and transported by air.The filters were balanced and weighted in a superclean laboratory with controlled temperature (20 ± 1°C) and RH (40 ± 3%), both before and after the sampling.

Water-Soluble Ion Measurement
Samples on teflon filters were firstly extracted ultrasonically using 10 mL ultrapure water (18.5 MΩ cm −1 ) for 30 minutes.The aqueous extract was filtered through a 0.45 μm water filter and the ion concentrations were determined using ion chromatography (Dionex, ICS 2000).A Dionex separator column of AS11-HC with KOH eluent was used for anion analysis (NO3 − , SO4 2− , and Cl − ), and a cation analytical column of CS12A and an eluent of 20 mM methyl sulfonic acid was used to analyze inorganic cations (Na + , NH4 + , K + , Mg 2+ , Ca 2+ ).Careful quality assurance and quality control (QA/QC) procedure was applied.Reference materials from the National Institute of Metrology, China were used as standards.Blank and standard samples were repeated every ten samples.Examples of a calibration curve of standard samples are displayed in the Supplementary Material.

Concentrations of PM10, PM2.5, and WSIIs
Table 1 summarizes the observed particle and WSII concentrations in the three sampling sites.During 2012-2013, the annual average PM10 and PM2.5 concentrations were 125.8 ± 74.4 and 86.7 ± 49.7 μg m −3 in CD, 116.3 ± 54.7 and 78.6 ± 36.8 μg m −3 in NJ, and 101.0 ± 51.7 and 71.7 ± 36.9 μg m −3 in CQ, respectively.Obviously, they all exceeded the latest NAAQS-II issued in 2012, and the annual PM2.5 concentrations were more than 2 times the 35 μg m −3 limit.Daily PM2.5 levels in more than one-third of the sampling days surpassed the daily 75 μg m −3 criteria, and there were 8, 3, and 3 heavy pollution days in CD, NJ, and CQ with PM2.5 higher than 150 μg m −3 .The average PM2.5/PM10 ratios were 0.72 in CD, 0.69 in NJ, and 0.71 in CQ, indicating a predominance of fine particulate pollution in the Sichuan Basin.
The annual averages of total WSIIs were 43.0 ± 27.9, 36.2 ± 18.4, and 35.4 ± 18.4 μg m −3 in CD, NJ, and CQ (Table 1), which accounted for 49.9%, 46.1%, and 49.4% of the PM2.5 mass, respectively.Although the average WSIIs in PM2.5 showed minor differences among sites, there were large fluctuations within each site, from a few to a couple of hundred μg m −3 .For instance, WSII levels in CD were lower than 10 μg m −3 during clean period (PM2.5 < 35 μg m −3 ), and they increased to higher During 2012-2013, an intensive haze campaign was carried out in the region and detailed information about fine particle pollution was gathered from Chengdu, Chongqing, and Neijiang (a medium-sized city in the Sichuan Basin).The one-year field sampling data is summarized, here, to provide a comprehensive study of the WSIIs in PM 2.5 of the Sichuan Basin.The WSII property is analyzed from annual and seasonal perspectives, and the acidity characteristics of the sulfate-nitrate-ammonium system and nitrate formation are investigated in the three cities of different sizes.The indicators reflected from the WSIIs are also explored to track the sources of PM 2.5 in the region.

Site Description and Field Sampling
Locations of the three sampling sites in the Sichuan Basin are denoted in Figure 1b.Chengdu is the provincial capital of Sichuan Province and surrounded by many small and medium-sized cities in the west of Sichuan basin.The sampling site in Chengdu is on the roof of a sixth-floor building with a height of 28 m (CD, 104 • 6 E, 30 • 36 N), which stands beside a main road with high traffic density.The site is representative of urban air quality, combining the influence of local vehicular emission, residential emission and regional pollution.
Chongqing is a municipality that is directly administrated by the Central Government.The sampling site in Chongqing is on the roof of a commercial building (CQ, 29 • 37 N, 106 • 30 E) in Yubei District in a downtown region (Figure 1b), and the sampling height is 35 m.The site is surrounded by main roads and office buildings.The third sampling site in Neijiang is on the roof of Neijiang Environmental Monitoring Center (NJ, 105 • 4 E, 29 • 42 N) with a height of 25 m.Neijiang is located 150 km southeast of Chengdu and 145 km west of Chongqing with an area of 75 km 2 and a population of about 1 million in the downtown region.
During May 2012 to May 2013, particulate samples were synchronously collected at the above three sites.Both PM 2.5 and PM 10 were sampled once every six days on 47 mm teflon filters using a four-channel sampler at a flow rate of 16.7 L/min (model: TH-16A, Tianhong Instrument Co., Ltd., Wuhan).The sampling process was carried out by trained staff of a local monitoring station, and the flow rate of the sampler was checked monthly using a bubble flow meter (Gilian Gilibrator 2, Sensidyne, US) to ensure collection efficiency.The sampled filters were stored in refrigerator at −18 • C and transported by air.The filters were balanced and weighted in a superclean laboratory with controlled temperature (20 ± 1 • C) and RH (40 ± 3%), both before and after the sampling.

Water-Soluble Ion Measurement
Samples on teflon filters were firstly extracted ultrasonically using 10 mL ultrapure water (18.5 MΩ cm −1 ) for 30 minutes.The aqueous extract was filtered through a 0.45 µm water filter and the ion concentrations were determined using ion chromatography (Dionex, ICS 2000).A Dionex separator column of AS11-HC with KOH eluent was used for anion analysis (NO 3 − , SO 4 2− , and Cl − ), and a cation analytical column of CS12A and an eluent of 20 mM methyl sulfonic acid was used to analyze inorganic cations (Na + , NH 4 + , K + , Mg 2+ , Ca 2+ ).Careful quality assurance and quality control (QA/QC) procedure was applied.Reference materials from the National Institute of Metrology, China were used as standards.Blank and standard samples were repeated every ten samples.Examples of a calibration curve of standard samples are displayed in the Supplementary Material.

Results and Discussion
3.1.Concentrations of PM 10 , PM 2.5 , and WSIIs Table 1 summarizes the observed particle and WSII concentrations in the three sampling sites.During 2012-2013, the annual average PM 10 and PM 2.5 concentrations were 125.8 ± 74.4 and 86.7 ± 49.7 µg m −3 in CD, 116.3 ± 54.7 and 78.6 ± 36.8 µg m −3 in NJ, and 101.0 ± 51.7 and 71.7 ± 36.9 µg m −3 in CQ, respectively.Obviously, they all exceeded the latest NAAQS-II issued in 2012, and the annual PM 2.5 concentrations were more than 2 times the 35 µg m −3 limit.Daily PM 2.5 levels in more than one-third of the sampling days surpassed the daily 75 µg m −3 criteria, and there were 8, 3, and 3 heavy pollution days in CD, NJ, and CQ with PM 2.5 higher than 150 µg m −3 .The average PM 2.5 /PM 10 ratios were 0.72 in CD, 0.69 in NJ, and 0.71 in CQ, indicating a predominance of fine particulate pollution in the Sichuan Basin.The annual averages of total WSIIs were 43.0 ± 27.9, 36.2 ± 18.4, and 35.4 ± 18.4 µg m −3 in CD, NJ, and CQ (Table 1), which accounted for 49.9%, 46.1%, and 49.4% of the PM 2.5 mass, respectively.Although the average WSIIs in PM 2.5 showed minor differences among sites, there were large fluctuations within each site, from a few to a couple of hundred µg m −3 .For instance, WSII levels in CD were lower than 10 µg m −3 during clean period (PM 2.5 < 35 µg m −3 ), and they increased to higher than 90 µg m −3 under heavy polluted period (PM 2.5 > 150 µg m −3 ).SO 4 2− was the most abundant species of WSIIs, with an average concentration of 17.7 ± 11.2 µg m −3 in CD, 18.1 ± 10.0 µg m −3 in NJ, and 17.6 ± 9.6 µg m −3 in CQ.Annual average concentrations of the other ions were ranked in the order of NO 3 − > NH 4 + > Cl − > K + > Na + > Ca 2+ > Mg 2+ in CD, whereas the order was NH 4 + > NO 3 − > K + > Cl − > Ca 2+ > Na + > Mg 2+ in both NJ and CQ (Table 1).The secondary inorganic components, in total, constituted about 90% of the total WSIIs (89.3% in CD, 92.4% in NJ, and 94.2% in CQ), and the rest of the ions each had a minor contribution.Among the three cities, PM 2.5 and WSII levels were highest in CD, followed by NJ and CQ.The three cities have roughly the same SO 4 2− levels, while CD was characterized by higher NO 3 − and NH 4 + concentrations, indicating that the sampling site in CD was more affected by motor vehicles from its surrounding road.Specifically, concentration of chloride was also the highest in CD, which might be associated with coal combustion.CD and NJ suffered from higher loadings of K + (Table 1), and it was a diagnostic tracer for intensive biomass burning in the suburban regions [17].[15].However, there was still an increase in NO 3 − (10.9 µg m −3 ) [15], which further highlights the importance of vehicular emissions in Chongqing.
Results of the study were also compared with previous measurements conducted in other cities of China.CD displayed lower loadings of PM 2.5 and SO 4 2− than Deyang, which was a medium-sized city in the Sichuan Basin located on the diffusion air pathways of Chengdu [19].WSII levels in the Sichuan Basin were generally lower than cities in northern China, i.e., Handan (2013-2014) [20], Shijiazhuang (2009)(2010) [21], and Taiyuan (2009-2010) [22].However, the SO 4 2− and NH 4 + levels were higher than those of Beijing during 2009-2010 [23], despite lower PM 2.5 and NO 3 − levels.When compared with cities in southern China, like Shanghai and the Pearl River Delta (PRD) [24], the pollution situation of PM 2.5 and WSIIs was more serious in the Sichuan Basin.

Seasonal Variations of PM 2.5 and WSIIs
The seasonal variations of PM 2.5 and WSIIs at the three sites are depicted in Figure 2. Notably, winter has the highest PM 2.5 levels (108.1, 97.6, and 97.5 µg m −3 in CD, NJ, and CQ, respectively), and was the most heavily polluted season in the Sichuan Basin.Autumn in CD and NJ also recorded high PM  The seasonal patterns of WSIIs were attributable to local and regional source variations between seasons, as well as meteorological factors, which affected their formation, transformation, and transport.The seasonal variations of NH4 + followed the changes of PM2.5 in each city, and NH4 + levels were highest in winter.Undoubtedly, the sulfate concentrations were also high in winter, which were due to poor dispersion in the cold season, enhanced in-cloud process under high RH, and long contact time for gas-liquid reaction under stable meteorology.Specifically, NJ and CQ suffered from the highest loading of sulfate in summer, though the PM2.5 concentrations were low.As a typical secondary ion, SO4 2− formation via homogeneous gas phase reaction was greatly enhanced at high temperature and intensive solar radiation in summer.It is worthy to note that autumn samples in CD showed an elevated level of sulfate with respect to summer (Figure 2, 17.9 μg m −3 in summer vs. 19.5 μg m −3 in autumn), which might be ascribed to the elevated PM2.5 loading in the season.
Different from sulfate, the season pattern of nitrate was characterized by winter maxima, autumn medium, and spring/summer minima at the three sites (Figure 2).Temperature and relative humidity are two important meteorological factors influencing the thermodynamic features of nitrate, and high temperature and low RH is highly favorable for nitrate volatilization [25].Therefore, the low temperature and high RH in winter and autumn are beneficial for nitrate stabilization.Moreover, the high loadings of PM2.5 in winter provided more aerosol surfaces for heterogeneous formation of nitrate [26].

Stoichiometric Analysis of Cations and Anions
To examine the ion balance and acidity of PM2.5 samples, the ion mass concentrations (μg m −3 ) are converted into microequivalents (μmol m −3 ) by the following equations.(

AE anion equivalent
Figure 3 illustrates the scatter plots of AE vs. CE in four seasons of CD, NJ, and CQ.Strong correlations between anion and cation equivalents were found for all the three cities (~1.0),The seasonal patterns of WSIIs were attributable to local and regional source variations between seasons, as well as meteorological factors, which affected their formation, transformation, and transport.The seasonal variations of NH 4 + followed the changes of PM 2.5 in each city, and NH 4 + levels were highest in winter.Undoubtedly, the sulfate concentrations were also high in winter, which were due to poor dispersion in the cold season, enhanced in-cloud process under high RH, and long contact time for gas-liquid reaction under stable meteorology.Specifically, NJ and CQ suffered from the highest loading of sulfate in summer, though the PM 2.5 concentrations were low.As a typical secondary ion, SO 4 2− formation via homogeneous gas phase reaction was greatly enhanced at high temperature and intensive solar radiation in summer.It is worthy to note that autumn samples in CD showed an elevated level of sulfate with respect to summer (Figure 2, 17.9 µg m −3 in summer vs. 19.5 µg m −3 in autumn), which might be ascribed to the elevated PM 2.5 loading in the season.Different from sulfate, the season pattern of nitrate was characterized by winter maxima, autumn medium, and spring/summer minima at the three sites (Figure 2).Temperature and relative humidity are two important meteorological factors influencing the thermodynamic features of nitrate, and high temperature and low RH is highly favorable for nitrate volatilization [25].Therefore, the low temperature and high RH in winter and autumn are beneficial for nitrate stabilization.Moreover, the high loadings of PM 2.5 in winter provided more aerosol surfaces for heterogeneous formation of nitrate [26].

Stoichiometric Analysis of Cations and Anions
To examine the ion balance and acidity of PM 2.5 samples, the ion mass concentrations (µg m −3 ) are converted into microequivalents (µmol m −3 ) by the following equations.

AE(anion equivalent
Figure 3 illustrates the scatter plots of AE vs. CE in four seasons of CD, NJ, and CQ.Strong correlations between anion and cation equivalents were found for all the three cities (~1.0), supporting that the measured eight ions were dominant species in the PM 2.5 ionic components.In CD, most of the samples in autumn and winter were above the 1:1 (AE/CE) line, indicating an acidic feature.By contrast, the majority of the samples in spring fall below the 1:1 line, demonstrating a deficiency of anions which might be associated with more alkaline dust particles.CO 3 2− and HCO 3 − were not measured by the method, and also contributed to the anion deficiency.In the summer of CD, most of the samples generally showed a balance between anions and cations, while some of them also denoted acidic features which might result from the enhanced formation of sulfate and loss of cations from the volatilization of nitrate and ammonium.Similar seasonal patterns of PM 2.5 acidity were also observed in NJ and CQ (Figure 3b, Figure 3c).Interestingly, it was reflected from the scatter plots (Figure 3) that aerosol acidity increased with the level of AE, indicating PM 2.5 samples under heavy pollution were mostly acidic.Tian et al. [15] also found increased acidity with aerosol pollution level.High humidity and low wind speed were common meteorological conditions for heavy pollution [27].They were unfavorable for horizontal dispersion and vertical mixing of pollutants but beneficial for the formation of nitrate and sulfate, therefore resulting in the acidic feature.
Atmosphere 2019, 10, x FOR PEER REVIEW 6 of 13 were not measured by the method, and also contributed to the anion deficiency.In the summer of CD, most of the samples generally showed a balance between anions and cations, while some of them also denoted acidic features which might result from the enhanced formation of sulfate and loss of cations from the volatilization of nitrate and ammonium.Similar seasonal patterns of PM2.5 acidity were also observed in NJ and CQ (Figure 3b, Figure 3c).Interestingly, it was reflected from the scatter plots (Figure 3) that aerosol acidity increased with the level of AE, indicating PM2.5 samples under heavy pollution were mostly acidic.Tian et al. [15] also found increased acidity with aerosol pollution level.High humidity and low wind speed were common meteorological conditions for heavy pollution [27].They were unfavorable for horizontal dispersion and vertical mixing of pollutants but beneficial for the formation of nitrate and sulfate, therefore resulting in the acidic feature.

Chemical Forms of Nitrate and Sulfate
The scatter plots of NH4 + vs. SO4 2− , SO4 2− + NO3 − , and SO4 2− + NO3 − + Cl − (all the above denote electron equivalent concentrations) in CD, NJ, and CQ are depicted in Figure 4.As (NH4)2SO4 is less volatile and preferentially formed compared to NH4NO3 and NH4Cl, the relationships between NH4 + and SO4 2− are firstly explored to investigate the chemical forms of sulfate and nitrate.It is reflected from Figure 4a-c that NH4 + was closely related with SO4 2− , and the data were mostly above the 1:1 (NH4 + /SO4 2− ) line, suggesting the complete neutralization of SO4 2− by NH4 + , and (NH4)2SO4 was the major species.However, there were a few exceptions below the 1:1 (NH4 + /SO4 2− ) line in the summer of NJ and CQ (Figure 4a-c).It was understandable that the formation of SO4 2− was greatly enhanced under the high temperature of summer while NH4 + was more easily removed by decomposition.Therefore, the samples did not have sufficient NH4 + to fully neutralize SO4 2− , and NH4HSO4 existed in summer.When it came to nitrate, the samples in spring mostly occupied enough NH4 + to neutralize both SO4 2− and NO3 − (Figure 4d-f) and formed (NH4)2SO4 and NH4NO3, in spite of a few exceptions.In other seasons, NH4 + was able to neutralize the secondary anions when their levels were low (SO4 2− + NO3 − < 0.5 μmol m −3 ) (Figure 4d-f).In fact, the abundance of NH4 + almost equaled to the sum of SO4 2− , NO3 − , and Cl − under low PM loadings (Figure 4g-i), and dominant anions existed in the form of (NH4)2SO4, NH4NO3, and NH4Cl.However, under high SO4 2− and NO3 − levels (SO4 2− + NO3 − > 0.5 μmol m −3 ), NH4 + was far from fully neutralized (Figure 4d-f).It was observed in most previous studies that high NO3 − levels were associated with high levels of NH4 + [28].In contrast, the relatively high NO3 − observed in the present study was with moderate levels of NH4 + , suggesting that the formation rate of nitrate may be much higher than other ions.The high NO3 − might also be associated with high levels of NO2 under heavy pollution.
The correlation coefficients between NO3 − and other cations in PM2.5 were further calculated in Table 2 to identify the chemical forms of nitrate.Na + and K + were found to be correlated with NO3 − in most seasons, and NaNO3 and KNO3 were also the major chemical species in aerosol particles.Notably, there were exceptionally high levels of NO3 − in the winter of CD, and NO3 − were significantly correlated with all cations and existed in various forms of NH4NO3, NaNO3, KNO3, Mg(NO3)2, and Ca(NO3)2.that the formation rate of nitrate may be much higher than other ions.The high NO 3 − might also be associated with high levels of NO 2 under heavy pollution.
The correlation coefficients between NO 3 − and other cations in PM 2.5 were further calculated in Table 2 to identify the chemical forms of nitrate.Na + and K + were found to be correlated with NO 3 − in most seasons, and NaNO 3 and KNO 3 were also the major chemical species in aerosol particles.Notably, there were exceptionally high levels of NO 3 − in the winter of CD, and NO 3 − were significantly correlated with all cations and existed in various forms of NH 4 NO 3 , NaNO 3 , KNO 3 , Mg(NO 3 ) 2 , and Ca(NO 3 ) 2 .

Formation Mechanism of Nitrate and Sulfate
Sulfur oxidation ratio (SOR), defined as n-SO 4 2− /(n-SO 2 + n-SO 4 2− ), and nitrogen oxidation ratio (NOR) defined as n-NO 3 − /(n-NO 2 + n-NO 3 − ), in CD, NJ, and CQ, are listed in Table 3 and used to indicate the secondary transformation processes.Generally, the SOR values were much higher than 0.10 (Table 3), demonstrating the occurrence of strong secondary oxidation of SO 2 to SO 4 2− throughout the year [29].The interseasonal variation of SOR peaked in summer in both CD and CQ, and it was explicable by the accelerated homogenous gas-phase oxidation of SO 2 under high temperature [30].Moreover, the increased oxidizing capacity from the greater production of ozone in summer also promoted SO 4 2− formation.By contrast, the highest SOR in NJ appeared in autumn instead of summer, and a good correlation was found between SOR and RH (r = 0.58).This suggested that high RH also increased the formation of SO 4 2− by prompting SO 2 oxidation through heterogeneous reaction [31], i.e., metal-catalyzed H 2 O 2 /O 3 oxidation and in-cloud process.The good correlation between SOR and RH in winter of CD (r = 0.66) and CQ (r = 0.71) also confirmed the existence of heterogeneous reaction.On days with elevated RH, the hygroscopic growth of sulfate would increase the liquid water content, and the aqueous phase on the aerosol surface could provide heterogeneous transformation vectors for gaseous pollutants (SO 2 ).Therefore, the elevated RH would largely promote the secondary formation of sulfate [32,33].The annual average NOR in CD, NJ, and CQ all surpassed 0.1 (Table 3), indicating the existence of secondary oxidation of NO 2 to NO 3 − [29].Reflected in Table 3, the NOR values had a different seasonal pattern from SOR, and reached their maxima in winter.Although the absolute concentrations of sulfate and nitrate both increased with PM 2.5 levels, their relative importance changed under different pollution levels.Table 4 listed the variations of NO 3 − /SO 4 2− ratios and NOR/SOR ratios as a function of different PM 2.5 levels.The continuous increase of NO 3 − /SO 4 2− ratio as a function of PM 2.5 concentrations (Table 4), as well as the increase of NOR/SOR ratio, indicated that NO 2 oxidation under heavy pollution was more significant than SO 2 , and nitrate formation might play an important role in haze in the Sichuan Basin.The above result is also supported by the findings of Hewitt [34] and Tian et al. [15].It is worth noting that the high concentrations of nitrate in the study were mostly collected during hazy and humid weather with high sulfate and acidity.Thus, the details of nitrate formation are discussed below.the study were mostly collected during hazy and humid weather with high sulfate and acidity.Thus, the details of nitrate formation are discussed below.Thus, these samples were categorized into NH4 + -rich conditions ([NH4 + ]/[SO4 2− ] ratio > 1.5).According to Pathak et al. [28], the relationship between [NO3 − ]/[SO4 2− ] and [NH4 + ]/[SO4 2− ] could be used to show the formation pathways of NO3 − .Under NH4 + -rich conditions, a linear relationship exists between them, suggesting homogeneous gas-phase formation for NO3 − , and, otherwise, hydrolysis of NOx on preexisting aerosols is responsible for the high NO3 − level [12,15,28].In Figure 5, the relative abundance of nitrate ([NO3 − ]/[SO4 2− ]) increased as the [NH4 + ]/[SO4 2− ] ratio increased (r = 0.81 in CD, r = 0.75 in NJ, and r = 0.68 in CQ), suggesting that nitrate formation via gas-phase reaction became evident in the NH3-H + -SO4 2− -H2O system in aerosol [35,36].In fact, the good relationship between the excess ammonium (excess [NH4 + ] = [NH4 + ] − 1.5[SO4 2− ]) and nitrate (Figure 6) further confirmed that the homogeneous gas-phase formation of nitrate was significant.Reflected from Figure 6, the increase of nitrate rate seemed to surpass the increase of excess [NH4 + ] under high concentrations (NO3 − > 0.25 μmol m −3 ).As more NO3 − led to more ions, it could further explain the observed high acidity of PM2.5 under high pollution levels in Section 3.3.However, it is hard to ignore that some plots are rather scattered in Figure 5, and the relationship between [NO3 − ]/[SO4 2− ] and [NH4 + ]/[SO4 2− ] were further explored under different acidity and RH (Figure 7).The scatter plots of [NO3 − ]/[SO4 2− ] and [NH4 + ]/[SO4 2− ] for both AE/CE > 1 and AE/CE < 1 displayed significant linear relationships, highlighting the importance of homogeneous gas-phase reaction.However, the correlation of R 2 = 0.45 between [NO3 − ]/[SO4 2− ] and [NH4 + ]/[SO4 2− ] for samples with RH > 75% was lower than samples with RH < 75% (R 2 = 0.65), and the plots were more scattered (Figure 7b).The result was consistent with other research in Suzhou and Chongqing [15,16], where they also tended to have better linear correlation under lower RH conditions (<75%).The scattered plots under RH > 75% implied the existence of different mechanisms other than the homogeneous reaction.The critical parameters to heterogeneous formation of nitrate via N2O5 hydrolysis on pre-existing particles include particulate hygroscopicity, surface area, and acidity.On the one hand, high RH relates to greater water content and surface areas of aerosols, which may   However, it is hard to ignore that some plots are rather scattered in Figure 5 for samples with RH > 75% was lower than samples with RH < 75% (R 2 = 0.65), and the plots were more scattered (Figure 7b).The result was consistent with other research in Suzhou and Chongqing [15,16], where they also tended to have better linear correlation under lower RH conditions (<75%).The scattered plots under RH > 75% implied the existence of different mechanisms other than the homogeneous reaction.The critical parameters to heterogeneous formation of nitrate via N 2 O 5 hydrolysis on pre-existing particles include particulate hygroscopicity, surface area, and acidity.On the one hand, high RH relates to greater water content and surface areas of aerosols, which may promote N 2 O 5 uptake on the aerosol surfaces.On the other hand, the high concentrations of PM 2.5 mass and large fractions of WSIIs with acidic conditions would favor the hydrolysis of N 2 O 5 [34].
Atmosphere 2019, 10, x FOR PEER REVIEW 10 of 13 promote N2O5 uptake on the aerosol surfaces.On the other hand, the high concentrations of PM2.5 mass and large fractions of WSIIs with acidic conditions would favor the hydrolysis of N2O5 [34].

Source Analysis of WSIIs
Principal component analysis (PCA) is applied in the study using SPSS version 16.0 software packages to explore the sources of WSIIs in CD, NJ, and CQ.In the analysis, all the WSIIs are considered as variables, and factors explaining more than 80% of the total variance are extracted.Varimax rotation is then used to redistribute the variance and provide a more interpretable pattern of the factors.Three factors in each city, with their component loadings, eigenvalues, and explained variance are displayed in Table 5.
Factor 1 in CD covered 61.1% of the total variance, and had high loadings of Na + (0.74), NH4 + (0.96), Cl − (0.77), SO4 2− (0.85), and NO3 − (0.93).NH4 + , SO4 2− , and NO3 − are typical secondary ions, and Cl − is a tracer for coal combustion.Thus, factor 1 in CD was recognized as a mixture of secondary aerosols and coal combustion, and the good correlation between Na + and Cl − indicated their similar origins or coexistence in aerosols.In factor 2, 14.9% of the total variance was explained and loadings of K + and Ca 2+ were much higher than other variables, indicating the contribution from biomass burning and natural dust.Factor 3 was responsible for 8.5% of total variance, and was heavily loaded by Mg 2+ (0.89).Mg 2+ was from construction dust, and the factor was related with the wide reconstruction work in urban Chengdu.

Source Analysis of WSIIs
Principal component analysis (PCA) is applied in the study using SPSS version 16.0 software packages to explore the sources of WSIIs in CD, NJ, and CQ.In the analysis, all the WSIIs are considered as variables, and factors explaining more than 80% of the total variance are extracted.Varimax rotation is then used to redistribute the variance and provide a more interpretable pattern of the factors.Three factors in each city, with their component loadings, eigenvalues, and explained variance are displayed in Table 5.
When compared with the observed values in 2011, the levels of WSIIs in Chengdu have shown a downward trend except for NO 3 − and Cl − [8].In Chongqing, the PM 2.5 and SO 4 2− levels have decreased by 57.3% and 31% compared to those in 2005-2006, whereas the NO 3 − level increased by 43% (7.8 vs. 5.46 µg m −3 ) [13].The similar trend of decreasing SO 4 2− and increasing NO 3 − in both Chengdu and Chongqing were related to the strict enforcement of desulfurization engineering and the soaring of the vehicular population in large cities of China [18].After 2013, the decrease of PM 2.5 concentration in Chongqing was minor and stabilized around 67.5 µg m −3 from 2015 to 2016 2.5 concentrations (101.6 µg m −3 in CD and 81.2 µg m −3 in NJ), whereas spring and summer were relatively clean (spring: 72.2 µg m −3 in CD and 67.5 µg m −3 in NJ; summer: 68.3 µg m −3 in CD and 68.7 µg m −3 in NJ).In CQ, PM 2.5 concentrations in spring (57.3 µg m −3 ), summer (64.9 µg m −3 ), and autumn (68.0 µg m −3 ) were rather close to each other.Atmosphere 2019, 10, x FOR PEER REVIEW 5 of 13 and 68.7 μg m −3 in NJ).In CQ, PM2.5 concentrations in spring (57.3 μg m −3 ), summer (64.9 μg m −3 ), and autumn (68.0 μg m −3 ) were rather close to each other.

Figure 2 .
Figure 2. Seasonal variations of PM 2.5 and WSIIs in CD, NJ, and CQ.

3. 4 .
Chemical Forms of Nitrate and Sulfate The scatter plots of NH 4 + vs. SO 4 2− , SO 4 2− + NO 3 − , and SO 4 2− + NO 3 − + Cl − (all the above denote electron equivalent concentrations) in CD, NJ, and CQ are depicted in Figure 4.As (NH 4 ) 2 SO 4 is less volatile and preferentially formed compared to NH 4 NO 3 and NH 4 Cl, the relationships between NH 4 + and SO 4 2− are firstly explored to investigate the chemical forms of sulfate and nitrate.It is reflected from Figure 4a-c that NH 4 + was closely related with SO 4 2− , and the data were mostly above the 1:1 (NH 4 + /SO 4 2− ) line, suggesting the complete neutralization of SO 4 2− by NH 4 + , and (NH 4 ) 2 SO 4was the major species.However, there were a few exceptions below the 1:1 (NH 4 + /SO 4 2− ) line in the summer of NJ and CQ (Figure4a-c).It was understandable that the formation of SO 4 2− was greatly enhanced under the high temperature of summer while NH 4 + was more easily removed by decomposition.Therefore, the samples did not have sufficient NH 4 + to fully neutralize SO 4 2− , and NH 4 HSO 4 existed in summer.

Figure 4 .
Figure 4. Scatter plots of ammonium and the major acidic anions in PM2.5 of (a, d, and g) CD, (b, e, and h) NJ, and (c, f, and i) CQ.

Figure 4 .
Figure 4. Scatter plots of ammonium and the major acidic anions in PM 2.5 of (a,d,g) CD, (b,e,h) NJ, and (c,f,i) CQ.When it came to nitrate, the samples in spring mostly occupied enough NH 4 + to neutralize both SO 4 2− and NO 3 − (Figure 4d-f) and formed (NH 4 ) 2 SO 4 and NH 4 NO 3 , in spite of a few exceptions.In other seasons, NH 4 + was able to neutralize the secondary anions when their levels were low (SO 4 2− + NO 3 − < 0.5 µmol m −3 ) (Figure 4d-f).In fact, the abundance of NH 4 + almost equaled to the sum of SO 4 2− , NO 3 − , and Cl − under low PM loadings (Figure 4g-i), and dominant anions existed in the form of (NH 4 ) 2 SO 4 , NH 4 NO 3 , and NH 4 Cl.However, under high SO 4 2− and NO 3 − levels (SO 4 2− + NO 3 − > 0.5 µmol m −3 ), NH 4 + was far from fully neutralized (Figure 4d-f).It was observed in most previous studies that high NO 3 − levels were associated with high levels of NH 4 + [28].In contrast, the relatively high NO 3 − observed in the present study was with moderate levels of NH 4 + , suggesting

Table 2 .
The correlation coefficients (R) between NO3 − and cations in PM2.5 of CD, NJ, and CQ.

Table 2 .
The correlation coefficients (R) between NO 3 − and cations in PM 2.5 of CD, NJ, and CQ.

Table 3 .
Seasonal variation of sulfur oxidation ratio (SOR) and nitrogen oxidation ratio (NOR) in CD, NJ, and CQ.

Table 5 .
PCA factor loadings of WSIIs in CD, NJ, and CQ. from the Ministry of Environmental Protection, China (NO.201009001) and the Funds for Excellent Teachers from Capital University of Economics and Business (2016).We also would like to thank the staff in Sichuan Provincial Environmental Monitoring Center, Neijiang Environmental Monitoring Station and Chongqing Environmental Monitoring Center for the help in the sample collection.