Using stable isotopes to trace sources and formation processes of sulfate aerosols from Beijing, China

Particulate pollution from anthropogenic and natural sources is a severe problem in China. Sulfur and oxygen isotopes of aerosol sulfate (δ34Ssulfate and δ18Osulfate) and water-soluble ions in aerosols collected from 2012 to 2014 in Beijing are being utilized to identify their sources and assess seasonal trends. The mean δ34S value of aerosol sulfate is similar to that of coal from North China, indicating that coal combustion is a significant contributor to atmospheric sulfate. The δ34Ssulfate and δ18Osulfate values are positively correlated and display an obvious seasonality (high in winter and low in summer). Although an influence of meteorological conditions to this seasonality in isotopic composition cannot be ruled out, the isotopic evidence suggests that the observed seasonality reflects temporal variations in the two main contributions to Beijing aerosol sulfate, notably biogenic sulfur emissions in the summer and the increasing coal consumption in winter. Our results clearly reveal that a reduction in the use of fossil fuels and the application of desulfurization technology will be important for effectively reducing sulfur emissions to the Beijing atmosphere.

The concentrations of WSII show seasonal changes (Fig. 3).
− NO 3 , F − and Ca 2+ concentrations are relatively high in spring and autumn compared to summer and winter. The − SO 4 2 concentration is slightly lower in winter than in summer. The concentrations of Cl − and Na + are much higher in winter than in summer, spring and autumn.
The ion balance as an indicator of the acidity of the aerosols was calculated using the ratios of the anion equivalents (AE) to the cation equivalents (CE) in TSP samples 2 (Fig. 3d). The ratios of AE/CE during the sampling period range from 0.38 to1.17 with a mean value of 0.83 ± 0.17 (n = 70). Most of ratios of AE/CE are lower than 1.0, which shows no seasonal changes.

Discussion
In order to determine the relationship between ions in TSP (as well as the sulfur and oxygen isotopic compositions), correlation coefficients are calculated (  2 ratio implies that mobile sources of the pollutants are predominant over stationary sources 42 . However, the majority of the ratios are lower than 0.8 during the heating period, suggesting the predominance of stationary sources (emission from coal combustion) over mobile sources of pollutants.
The Cl − and Na + concentrations are positively correlated (r = 0.83) and display an increase in winter (Fig. 3c), suggesting they have a common source. As the prevailing winds in Beijing's winter are north and northwest, a significant contribution from seawater in Beijing aerosols can be ruled out. In addition, the ratio of Cl − to Na + in winter is 3.2 ± 1.2, which is different from the ratio in seawater of 1.17 43 . Studies reported that high Cl − concentration in Beijing aerosols may be related to coal combustion 2,43 . During combustion, complex changes in coal particles may cause the vaporization of volatile elements, including sodium 44 . Sodium vaporised from coal during combustion, may be present in the gas phase or bound in particulate aerosols in the flue gases, which can be emitted to the atmosphere 44 . Significant correlations exist as well for δ 34 S and Cl − (r = 0.67) and for δ 34 S and Na + (r = 0.66), which provides additional evidence for a common origin and the significant contribution of coal combustion to the atmospheric sulfate pool.
Water-soluble sulfate in aerosol is derived from both primary (e.g. sea salt, dust, fly ash) and secondary (e.g. oxidation of SO 2 and H 2 S) sulfates 14,37 , all characterized by their own distinct isotopic composition. Consequently, the sulfur isotopic composition of sulfate in Beijing aerosols reveals a mixture from different sulfate sources with high and low δ 34 S values (Fig. 6). Volcanism as a source of sulfate in Beijing aerosols can be excluded with no volcanic activities in North China 45 . Also, a significant contribution from sea salt is not very likely as suggested by the low concentration of Na + (mean = 1.2 ± 1.0 μ g/m 3 , n = 70). In addition, the weak correlation between − SO 4 2 and Na + (r = 0.16) as well as − SO 4 2 and Cl − (r = 0.19) also suggest that seawater sulfate provides only a very small contribution to the aerosols in Beijing.
Some SO 2 emissions from industry and transportation ultimately originate from oil combustion. It has been estimated that 15.3 million tons of petroleum products were consumed in Beijing in 2012, including 4.4 million tons by industry and 6.1 million tons from transportation 46 . The sulfur content in oil from North China ranges from 0.1% to 0.6% and its δ 34 S value varies between 13.7‰ and 24.2‰ (mean = 20.5‰, n = 4) 47 . The emission rate of SO 2 from oil combustion is relatively constant with almost no seasonal change in the consumption of petroleum. Therefore, the SO 2 emissions from oil combustion in the study area are a steady source of sulfate in the aerosol that is characterized by a relatively high δ 34 S value.
The contribution from coal combustion to the atmospheric sulfur pool of China is very significant 15,37 . In Beijing, the total consumption of coal was 22.7 million tons in 2012 based on the China Energy Statistical Yearbook (2013) 46 . With an average sulfur content of 0.77wt.% in coal from North China 48 , an estimated 174.8 thousand tons of sulfur were released into Beijing's atmosphere in 2012 assuming that there is no desulfurization implemented into the coal combustion processes. During the winter, the consumption of coal in China is increasing due to heating, which will affect the overall δ 34 S value of atmospheric sulfate. It has been shown before that the sulfur isotopic composition of atmospheric sulfur in the different regions of China is closely related to the sulfur isotope signature of the coal used in the respective area 37 . Reported δ 34 S values for coal from North China (mean = + 6.6‰) are higher than for coal from South China (mean = − 0.32‰) 47,48 . The average δ 34 S value of sulfate in aerosol from Beijing (6.6 ± 1.8‰, n = 70), determined in this study, is similar to that for coal used in North China, indicating that coal combustion is a significant, if not the most important contributor to the atmospheric sulfate pool.
Biogenic sulfur from wetlands and soils is an important source for atmospheric sulfate, especially in the summer 22,23 , and δ 34 S values for biogenic sulfur are generally negative, ranging from − 10 to − 2‰ [21][22][23] . Low δ 34 S values recorded for Beijing summer aerosols may, thus, reflect a larger contribution from biogenic sulfur, in contrast to the winter season where low temperatures greatly attenuate (or inhibit) microbial activity in wetlands and soils.
Sulfates in aerosols can also originate from terrigenous sources (e.g. soil or mineral dusts). In this study, a significant correlation can be seen between − SO 4 2 and Ca 2+ concentrations (r = 0.63), which suggests a terrigenous  contribution to the aerosol sulfate pool. Ca 2+ as a reference element for mineral dust is used for calculating the proportion of this contribution (f ts ) to sulfate in the aerosol by following the equation 49 :  51,52 . Assuming that the seasonal change in the δ 34 S values of aerosol sulfate reflects temporal variations in the proportional ctributions from different sulfate sources, these proportions can be estimated using the following equations 16,20 : oc cc bs ts where f oc , f cc , f bs and f ts represent the fractional contributions of oil combustion, coal combustion, biogenic source and terrigenous source, respectively, and δ 34 S oc , δ 34 S cc , δ 34 S bs and δ 34 S ts represent the corresponding δ 34 S value of each sulfur source. In winter, biogenic sulfur is likely negligible since the soil microbial activity is weak. Hence, we assume that in winter, f bs equals 0. In addition, the contribution of oil combustion is relatively constant throughout the year as there is no seasonal variation in oil consumption. By solving equations (1)-(3), the contributions of sulfate sources can be calculated (results listed in Table 3), assuming a δ 34 S oc value of 20.5 ± 4.8‰ 47 , a δ 34 S cc value of 6.6 ± 3‰ 47,48 , a δ 34 S bs value of − 6 ± 4‰ 21-23 and a δ 34 S ts value of 4.5 ± 3.5‰ 51,52 as the respective δ 34 S signature of each sulfur source. The results show that the average contributions of coal combustion, oil combustion, biogenic sulfur and terrigenous sulfate to sulfate in aerosols of Beijing are 49.6 ± 7.5%, 17.6 ± 8.6%, 19.8 ± 9.9% and 10.1 ± 6.2%, respectively, but exhibiting strong seasonal differences (Table 3).
It has been shown that the seasonal change in the proportion of different oxidation pathways of atmospheric sulfur dioxide to aerosol sulfate may also lead to a seasonality in δ 34 S sulfate 37,39 . The sulfur isotope fractionation factors (α ) for different oxidation reactions of SO 2 are distinct to each other [28][29][30][31] . Experimental studies show that the fractionation factor for heterogeneous oxidation is 1.0165 ± 0.001 at 25 °C 29,30 . The fractionation factor during gas-phase oxidation by the OH radical (homogeneous oxidation) is 0.991, which is determined by using an ab initio quantum mechanical calculation 31 . In contrast, results from laboratory measurements show that the fractionation factors during homogeneous oxidation and aqueous oxidation by H 2 O 2 and O 3 are greater than 1.0, while the fractionation factor for oxidation by transition metal ion catalysis (TMI-catalysis) is α Fe = 0.9894 ± 0.0043 at 19 °C 28 . A recent study shows that the changing proportion among oxidation by TMI-catalysis, OH and H 2 O 2 was the main cause for the seasonality in the δ 34 S values of sulfate versus SO 2 39 . However, sulfate from aqueous SO 2 oxidation by TMI-catalysis only accounts for 9-17% of the global sulfate production 53 . Thus, it cannot resolve the difference in δ 34 S sulfate observed between summer and winter in Beijing aerosol sulfate.
A strong negative correlation between mean air temperature and δ 34 S sulfate in aerosol (r = − 0.83, Fig. 4b) is apparent. This could indicate that the seasonality in δ 34 S of atmospheric sulfate may result from a seasonal Na + NH 4    The maximum difference between the δ 34 S values in summer and winter, however, is 7.9‰, which is substantially higher. Consequently, the temperature effect alone cannot explain the seasonal difference in δ 34 S. In addition, a recent study shows that for the oxidation of SO 2 by OH radicals, H 2 O 2 and transition metal ion catalysis (TMI-catalysis), the coefficients of the temperature effect on the fractionation factors are 0.004 ± 0.015‰°C 1− , 0.085 ± 0.004‰°C 1− and 0.237 ± 0.004‰°C 1− , respectively 39 . This indicates that the temperature effect on the fractionation factor is negligible for the OH radical pathway, but can be more significant for the oxidation by TMI-catalysis and H 2 O 2 39 . A temperature difference of 30 °C between summer and winter could account for a maximum seasonal isotopic difference via TMI-catalysis of 1.2‰, again insufficient for the observed maximal seasonality in δ 34 S sulfate .
In addition to temperature, a positive correlation between atmospheric pressure and δ 34 S sulfate values can be observed (r = 0.69, Fig. 4b). Leung et al. evaluated the sulfur isotope fractionation factor (α ) during the oxidation of SO 2 by OH radicals based on the RRKM transition-state theory, and found that the factor is a function of pressure and temperature, i.e., α = 1.1646 + 0.0198(P/Torr) 0.1769 − 0.3092[(T/K)/1000] 32 . The maximum difference in atmospheric pressure between summer and winter is around 4 kpa (30 Torr), which may cause a change in the δ 34 S sulfate value by 0.45‰ for the OH oxidation of SO 2 . It suggests that the change in atmospheric pressure may have only a minor effect on the variation in δ 34 S sulfate .
Considering that changes in the sulfur isotopic fractionation resulting from meteorological boundary conditions (i.e. air temperature and atmospheric pressure) are insufficient to explain the observed seasonality in δ 34 S sulfate , respective variations are more likely reflecting temporal changes in the proportional contributions from different sulfate sources during different times of the year, as has been reported from other areas before 38 . It is proposed here that during the summer, biogenic sulfur emissions which are characterized by negative δ 34 S values, are an important source of atmospheric sulfate 22,23 , leading a decrease in the overall δ 34 S value of aerosol sulfate in the summer. In contrast, the increase in coal consumption for heating during winter time (and with it an increase in the proportional importance of this contribution to the overall sulfate pool) will cause a shift to a more positive overall δ 34 S value for aerosol sulfate in the winter.
Evidence in particular for the latter, i.e. the increasing coal combustion in the winter, comes from the oxygen isotopic composition of sulfate aerosol (δ 18 O sulfate ). It also exhibits strong seasonal changes, with the highest values in winter and low values in summer (Fig. 4a). Previous studies have shown that high-temperature combustion processes, thereby oxidizing the sulfur dioxide to sulfate, will lead to 18 O enriched aerosol sulfate 25,33 , and δ 18 O sulfate values of + 35 to + 40‰ have been reported 33 . Consequently, a higher contribution from coal combustion in the winter will cause more positive δ 18 O sulfate values of sulfate aerosols. In addition, the lack of a positive correlation between the δ 18 O sulfate and the δ 18 O H2O (Fig. 7) supports the assumption that sulfate formed at high temperatures, rather than heterogeneous, i.e. aqueous oxidation of SO 2 is the important process during the winter, because the latter would result in a positive correlation between δ 18 O sulfate and the δ 18 O H2O 55 . This explains the obvious decoupling of both oxygen isotope records (Fig. 7) and argues that the observed increase in δ 18 O sulfate seen in the winter is reflecting most likely a source effect, i.e. the high-temperature combustion of coal, generating an 18 O enriched primary sulfate aerosol.
Although an influence of the meteorological boundary conditions, i.e. air temperature and atmospheric pressure, on the observed seasonality in the sulfur and oxygen isotope compositions of sulfate in Beijing aerosol cannot be ruled out, the tight coupling of the temporal trend in δ 34 S sulfate and δ 18 O sulfate is best explained as a variation related to the source of aerosol sulfate. Due to their pronounced seasonality, the two strongest variables in this respect are contributions from biogenic sulfur emissions, limited to the summer, and increasing coal consumption in the winter. In particular in the winter, coal combustion is the main contributor to the Beijing aerosol  sulfate pool as evidenced by the paired shift to high δ 34 S and δ 18 O values for sulfate in aerosol. Moreover, the temporal variation in PM 2.5 concentration during the sampling period exhibits an obvious seasonal trend (Fig. 8) that is similar to the temporal records of aerosol sulfate sulfur isotopic composition and the maximum contribution from coal combustion to the aerosol sulfate pool, i.e. high in winter and low in summer. This strongly underlines the conclusion that coal combustion is the major contributor to the Beijing aerosol (sulfate) pool. While the biogenic sulfur emissions are likely more difficult to control, a reduction in the usage of coal, especially the high sulfur coal, and the application of desulfurization measures for coal powered industries will be important in reducing sulfur emissions to the Beijing atmosphere, which will ultimately improve Beijing's air quality.

Methods
Sample Collection. Total suspended particulates (TSP) were sampled on a 3-day basis from May 31, 2012 to June 10, 2014 (n = 73) on top of the roof (around 10 meters above ground level) of the Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing (Fig. 1). The samples were collected using a high volume air sampler (Qingdao Laoshan, KC1000) with a flow rate of 1.0 m 3 min −1 and pre-combusted (450 °C for 6 h) quartz fiber filters (20 cm × 25 cm, Pallflex). After sampling, a pre-combusted glass jar (150 ml) with Teflon lined screw cap was used to store each filter in a freezer at − 20 °C until geochemical analyses. The meteorological data during sampling were obtained from China Meteorological Data Sharing Service System (http://cdc.nmic.cn/home.do, Fig. 2). The daily average values of air temperature, air humidity, wind speed and atmospheric pressure were calculated based on the observation data at 2.00 a.m., 8.00 a.m., 14.00 p.m. and 20.00 p.m. The detection limits of air temperature, precipitation, air humidity, wind speed and atmospheric pressure are 0.1 °C, 0.1 mm, 1%, 0.1 m/s and 0.1 hPa, respectively. Analytical Methods. Using a circular hole-puncher, two circular pieces with a diameter of 47 mm were cut from each filter (20 cm × 25 cm), shredded and soaked in 200 ml of Milli-Q water for 30 minutes added by ultrasonification 14 . Subsequently, the filters were kept in water overnight in order to quantitatively extract the water-soluble ions. The quartz filter fibers were removed by filtration using 0.45 mm millipore filters. 10 ml of the solution were taken for ion concentration analysis, and the remaining solution was acidified to pH< 2 by addition of HCl solution and heated to boiling. The dissolved sulfate in the solution was precipitated as barite by adding 25 ml of 8.5% BaCl 2 solution and the glass beaker with the solution was kept at 80 °C for additional 3 hours. The BaSO 4 precipitates were collected on 0.22 μ m acetate millipore filters and rinsed with 150 mL Milli-Q water to remove Cl −15 . The millipore filters with the precipitates were dried in an oven at 45 °C for 48 hours. Then the S and O isotopic compositions of the BaSO 4 precipitates were analyzed. The blank samples were also analyzed after the same method.
The concentrations of the water-soluble ions ( − SO 4 2 , − NO 3 , Cl − , F − , + NH 4 , Na + , K + , Ca 2+ and Mg 2+ ) were analyzed by ion chromatography (Dionex ICS900). An IonPac TM AS23 column (4 × 250 mm, Dionex) and an IonPac TM CS12A column (4 × 250 mm, Dionex) were used f the determination of anions and cations, respectively. The 4.5 mM sodium carbonate and 0.8 mM sodium bicarbonate were used as the eluent for anions; the 20 mM methansulfonic acid (MSA) was used as the eluent for cations. The detection limits were below 0.07 μ g/m 3 for cations and anions.