Particle Size Distributions of PCDD / Fs and PBDD / Fs in Ambient Air in a Suburban Area in Beijing , China

The particle size distributions of polychlorinated and polybrominated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs and PBDD/Fs, respectively; together labeled PXDD/Fs) in ambient air in a suburban area in Beijing, China, were determined. The sums of the concentrations of the 17 2,3,7,8-PCDD/Fs and the sums of the concentrations of the 13 2,3,7,8-PBDD/Fs that were analyzed were 1499–2799 fg m (95.4–175.4 fg I-TEQ m) and 1171–2424 fg m (42.2–109.3 fg TEQ m), respectively. The PXDD/Fs were mainly (~90%) in the particulate phase. Significant linear correlations were found between the gas/particle partition coefficients (Kp) and subcooled liquid vapor pressures (PL) of the PXDD/Fs. The regression coefficients indicated that the PCDD/Fs were mainly adsorbed to the particles and that the PBDD/Fs were mainly absorbed by the particles. The concentrations of the PXDD/Fs increased as the particle size decreased. The highest PXDD/F concentrations were found in the dae < 1.0 μm particles and more than 80% of the PXDD/Fs were found to be in the dae < 2.5 μm particles. Similar regression coefficients were found for the KP against PL for the different particle size fractions in the air. The PXDD/F distribution profiles in particles of different sizes were also studied. The lower chlorinated PCDD/Fs were found at higher concentrations in the coarser particles, and the higher chlorinated PCDD/Fs were mainly found in the finer particles. Polybrominated dibenzofurans, particularly the higher brominated dibenzofurans, were the dominant PBDD/F congeners. The contributions of the higher brominated dibenzofurans to the total PBDD/F concentrations decreased as the particle size increased, but that was not the case for the polybrominated dibenzo-p-dioxins.


INTRODUCTION
Polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) are highly toxic pollutants that are ubiquitous in the air.PCDD/F concentrations in the air in China have increased over the past two decades (Zhao et al., 2011;Liu et al., 2013).Polybrominated dibenzo-p-dioxins and dibenzofurans (PBDD/Fs) have been found to have similar physicochemical properties, toxicities, and geochemical behaviors in the environment to PCDD/Fs (Melber and Kielhorn, 1998;Birnbaum et al., 2003;Ashizuka et al., 2005).Exposure to PCDD/Fs and PBDD/Fs (together called PXDD/Fs) leads to risks to human health because PXDD/F bioaccumulate and cause many different cardiopulmonary diseases (Vena et al., 1998;Tanabe, 2002;Safe, 2003).
Ambient air is one of the most important environmental media through which PXDD/Fs are transported.PXDD/Fs transported in ambient air can be deposited through several pathways (Mi et al., 2012;Tseng et al., 2014a, b;Chandra Suryani et al., 2015).The PXDD/F distribution in ambient air is influenced by a number of factors, such as the temperature, humidity, properties of the suspended particles, and saturated vapor pressures of the PXDD/F congeners (Martínez et al., 2006).Determining the distributions of pollutants in particles of different sizes in ambient air can help in identifying the origins of the pollutants and in determining the behaviors of the pollutants in the atmospheric environment.A number of studies have been performed in different parts of the world aimed at determining the particle size distributions of certain airborne pollutants, including polycyclic aromatic hydrocarbons (Allen et al., 1996;Kiss et al., 1998;Duan et al., 2007), polybrominated diphenyl ethers (Deng et al., 2007;Mandalakis et al., 2009), organochlorine pesticides and polychlorinated biphenyls (Chrysikou et al., 2009;Landlová et al., 2014).The particle size distributions of airborne PCDD/F congeners have been found to vary considerably in a number of studies (Kaupp et al., 1994;Kaupp and McLachlan, 2000;Oh et al., 2002;Chao et al., 2003).It is therefore difficult to describe the general particle size distribution of airborne PCDD/Fs.The PBDD/Fs can be compared with the PCDD/Fs, but less concern has been expressed about the presence of PBDD/Fs in the environment than about the presence of PCDD/Fs, and very little information is available on PBDD/F concentrations in the environment.Previously published studies of PBDD/Fs in the environment have mainly been focused on the sources of PBDD/F emissions, such as electronic waste incineration processes (Duan et al., 2011), municipal waste incinerators (Wyrzykowska et al., 2009;Gullett et al., 2010) and metallurgical processes (Wang et al., 2010;Li et al., 2015b;Wang et al., 2015).Few studies of PBDD/Fs in the ambient air have been carried out in recent years (Li et al., 2008a;Li et al., 2011a).Very few studies of the particle size distributions of PBDD/Fs in ambient air have been performed (Hayakawa et al., 2004).Data on the particle size distributions of PBDD/Fs in ambient air are limited because of the low PBDD/F concentrations that are found in each particle size fraction.PBDD/Fs are often not included in monitoring programs, and this may be another reason for few data on the particle size distributions of PBDD/Fs in ambient air being available.However, it is important for PBDD/Fs in air to be studied so that the origins and transportation of PBDD/Fs in the atmosphere can be better understood.There is therefore a significant and urgent need for the particle size distributions of airborne PXDD/Fs to be studied.
Beijing is one of the largest cities in China.Rapid economic growth, increasing traffic volumes, and an increasing population in the last two decades have led to Beijing suffering a great number of air pollution problems (Hao and Wang, 2005).It is worth noting that haze weather has been occurring more frequently in recent years than previously.Understanding the particle size distributions of airborne PXDD/Fs could allow us to improve our understanding of the risks posed by haze to public health, and could improve our ability to control such risks.The partitioning of atmospheric PXDD/Fs to particles of different sizes in ambient air in a suburban part of Beijing was investigated in the work presented here, and the PXDD/F homolog and congener profiles in particles of different sizes were assessed.

Air Sampling
Air samples were collected at the Beijing urban ecosystem research station (116°12′E, 40°00′N), which is in the northwest suburbs of Beijing City.The sampling site is not close to any main roads or obvious sources of pollution.The samples were collected using a KS-303 PM 10/2.5/1.0 sampler (Kálmán System, Hungary).The sampler was placed on the roof of a four-floor building (~12 m high).
The sampler was equipped with four quartz-fiber filters (QFFs; Ahlstrom Munktell, Sweden), which captured particles with aerodynamic diameters (d ae s) of > 10 µm, 2.5-10 µm, 1.0-2.5 µm, and < 1.0 µm, to collect particlebound PXDD/Fs.After the filters the sampler contained a glass cartridge containing polyurethane foam (PUF; 63 mm in diameter, 76 mm long; Tisch Environmental, USA) for collecting gaseous chemicals.Each QFF was baked at 450°C for 12 h before use, to remove organic contaminants.Each QFF was weighed before and after sampling, to determine the mass of particles that was collected.Before use, each PUF sampler was extracted with acetone for 45 min then with a 1:1 v/v mixture of dichloromethane and hexane for another 45 min in an accelerated solvent extraction system (Thermo Fisher Scientific, USA).Each PUF cartridge was then dried under vacuum in a desiccator, then stored in a sealed polyethylene bag.The sampling flow rate was 400 L min -1 throughout each sampling period.After a sample had been collected, the QFFs and PUF were wrapped separately in hexane-rinsed aluminum foil, to protect them from light, and each was stored in a sealed polyethylene bag.The samples were then transported to the laboratory and stored at -18°C until they were analyzed.The sampling procedure details are shown in Table 1.

Analysis
Each filter and PUF was spiked with 13 C 12 -labeled PCDD/F (EPA-1613LCS, Wellington Laboratories, Canada) and PBDD/F (EDF-5408, Cambridge Isotope Laboratories, USA) surrogate standards and then extracted with a 1:1 v/v mixture of dichloromethane and hexane using an accelerated solvent extraction system.Each extract was concentrated using a rotary evaporator and transferred into pure hexane.The extract was then cleaned using a multilayer silica gel column, a basic alumina column, and then an activated carbon column.The clean-up procedure has been described in detail previously (Li et al., 2015a).The extract was then concentrated under a gentle stream of nitrogen, and transferred into 10 µL of nonane in a mini-vial. 13C 12 -labeled injection standards (EPA-1613 IS, EDF-5409) were then added to allow the recoveries of the surrogate standards to be determined, and the vial was vortexed to completely mix the injection standards and samples.
The extracts were analyzed for PCDD/Fs using a gas chromatograph coupled with an Autospec Ultima highresolution mass spectrometer (Waters, USA) using an electron impact ionization source.The analytical method that was used has been described in detail previously (Li et al., 2011b).The PBDD/Fs were quantified on Thermo Trace 1310 gas chromatography coupled with a Double Focusing Sector mass spectrometer spectrometer (DFS-MS) with an electron impact (EI) ion source.The high resolution mass spectrometer (HRMS) was operated in MID mode at RX10,000.Exactly, 1 µL of sample solution was injected with a CTC PAL autosampler in splitless mode into a 15 m (length) × 0.25 mm (inside diameter) × 0.1 μm (thickness) DB-5 MS fused silica capillary column.Helium served as the carrier gas with a constant flow of 1.0 mL min -1 .The electron emission energy was set to 45 eV, and the source temperature was 280°C.The oven temperature programs were employed as follows: start 120°C held for 1 min, 120-220°C at 12 min -1 , 220-260°C at 4°C min -1 , 260-320°C at 3 °C min -1 held for 7 min.The monitored traces of 2,3,7,8-sbstituted PBDD/F congeners were given in the supplementary material Table S3.

Quality Assurance and Quality Control
A breakthrough test, adding a half PUF cartridge in series after the first PUF, was performed, and no breakthrough was found after a 24 h sampling period.All of the samples were spiked with 13 C 12 -labeled compounds before extraction and at the end of the sample preparation procedure to allow the performance of the analytical method to be monitored.Five pairs of field blank and laboratory blank were analyzed.The signal/noise ratio > 3 was used to calculate the limits of detection (LOD) in this study.The concentrations of PXDD/Fs below the LOD were assigned concentrations equal to half the LOD.No differences were found between the field blank and laboratory blank in PXDD/F concentrations.The PCDD/F recoveries were 49-109% and the PBDD/F recoveries were 33-95%.

Particle Size Distributions
The mass concentrations of the particles of different sizes in the as-obtained samples are presented in Table 1.The total particle concentrations were 274-379 µg m -3 .The masses of particles collected on the back filters (d ae < 1.0 µm) were higher than the masses collected on the other filters, and the d ae < 2.5 µm size fraction contributed ~77% of the particle mass in each sample.

PCDD/Fs and PBDD/Fs in the Gas and Particle Phases
There were clear differences in the gas/particle partitioning behaviors of the different PXDD/F congeners, as shown in Fig. 1.A large fraction of each PXDD/F congener was found in the particulate phase.A larger fraction of each polychlorinated dibenzo-p-dioxin congener than of the equivalent polychlorinated dibenzofuran congener was associated with particles, and this would mainly have been because dibenzo-p-dioxin congeners have lower vapor pressures than the corresponding dibenzofuran congeners.Generally, the lower chlorinated PCDD/F congeners (tetra-, penta-CDD/Fs) usually tend to partition to the gas phase compared to the highly chlorinated ones (hexa-, hepta-, octa-CDD/Fs) at a specific temperature due to the higher saturation vapor pressure, which is similar to the previous reports (Lohmann et al., 2000;Li et al., 2008b).
The gas/particle distribution of a semi-volatile organic compound is typically described using the term K p , calculated as shown in Eq. (1): where TSP is the total particulate matter concentration (µg m -3 ), and F and A are the concentrations of the compound of interest in the particulate and gas phases, respectively.Plotting logK p against the logarithm of the subcooled liquid vapor pressure, log P L 0 , of the compound allows the partitioning constant to be calculated, as shown in Eq. ( 2) (Yamasaki et al., 1982): where m r is the slope of the trend line and b r is the y-intercept.P L 0 can be calculated using Eq.(3) (Hung et al., 2002): 1.67 10 ( ) 8.807 where RI is the gas chromatographic retention index of the compound of interest and T is the ambient temperature (K).
The slope m r should be close to -1 when true equilibrium partitioning occurs when either adsorption or absorption mechanisms are involved (Pankow, 1994).However, the slope m r can indicate whether adsorption or absorption is the dominant mechanism in the gas/particle partitioning process (Goss and Schwarzenbach, 1998).A plot of logK p against log P L 0 for our PCDD/F data is shown in Fig. 2. The vapor pressures of the PCDD/F congeners at the average ambient temperature of 17°C were calculated from the relationship between the P L 0 and the gas chromatographic retention indices.The PCDD/F retention indices from previous publications (Hale et al., 1985;Donnelly et al., 1987) were used.The P L 0 values for the PBDD/F congeners were calculated using previously published data (Gajewicz et al., 2010).A statistically significant linear correlation was Table 2. Parameters for the correlations between log K P and log P L 0 for the five total particle samples.R 2 is the square of the Pearson's product moment correlation coefficient, b r is the intercept, and m r is the slope of the trendline.found between the logK p and log P L 0 data for the PCDD/Fs, and the slope (-1.078) was < -1.This indicated that the partitioning of the PCDD/Fs between the particulate and gas phases in the air samples had not approached equilibrium, but was relatively close to equilibrium, possibly because traffic and other sources of pollution had relatively little influence on the sampling site.The slope also indicated that the PCDD/Fs were mainly adsorbed onto the particles, in agreement with the results of a previous study (Harner and Bidleman, 1998).The slopes for the PBDD/Fs were > -0.6, indicating that the Ʃ 13 2,3,7,8-PBDD/F congeners were mainly absorbed by the particles (see Table 2), unlike the PCDD/Fs.Kurokawa et al. (1998) found that the PCDD/F concentrations in particles of d ae < 1.1 µm contributed more than 50% of the total PCDD/F concentrations in their samples (Kurokawa et al., 1998).In another study, more than 60% of the PCDD/Fs were associated with particles of d ae < 0.41 µm, and 90% of the PCDD/Fs were found in particles of d ae < 2.1 µm (Oh et al., 2002).The PXDD/F concentrations in the particles of different sizes in our study are presented in Fig. 3.It was found that the PXDD/Fs were mainly in the particles of d ae < 1.0 µm (79.6% of the PCDD/Fs and 65.0% of the PBDD/Fs), and only small amounts of the PXDD/Fs were in the coarse particles (3.0% of the PCDD/Fs and 7.9% of the PBDD/Fs).The fraction of the total PXDD/F concentration in particles of each particular size decreased as the particle size increased, as was found in previous studies (Oh et al., 2002;Degrendele et al., 2014).In addition, it is noteworthy that the 12.3% of PCDD/Fs and 15.8% of PBDD/Fs were in the d ae 1.0-2.5 µm size fraction.In other words, the PBDD/F distribution was therefore similar to the PCDD/F distribution, both being strongly distributed to fine particles.In summary, more than 80%, on average, of the PXDD/Fs were in the particles of d ae < 2.5 µm (i.e., PM 2.5 ).

Relationships between the PCDD/F and PBDD/F Concentration Distributions and the Particle Size
We drew the following conclusions about the PXDD/F distributions from the results described above.Clearly different gas/particle phase concentrations were found for different PXDD/Fs, but large proportions of all of the PXDD/F congeners were found in the particulate phase.The concentration of the PXDD/F in the particulate phase as a proportion of the total PXDD/F concentration was negatively related to the particle size.
Another usual approach for studying the distribution of pollutants in the different size fractions is to use normalized histograms representing dC/dlogDp versus Dp, where dC is the concentration of PXDD/Fs in each size fraction; and Dp is the aerodynamic diameter (Chao et al., 2003;Chrysikou and Samara, 2009).Fig. 4 shows that the PXDD/Fs were mainly accumulated in the fine particle size fraction (d ae < 1.0 µm), might due to their physical and chemical properties and the higher surface area of the fine particles in the atmosphere.The distribution concentration of PBDD/Fs in the d ae 1.0-2.5 µm size fraction is relatively high and notable and should not be ignored.In addition, the calculated mass median diameters (MMDs) of PCDD/Fs and PBDD/Fs are 0.17 µm (geometric standard deviation, σ g < 0.01 µm) and 0.17 µm (σ g < 0.01 µm), respectively.

Relationship between the Gas/Particle Partitioning of the PXDD/Fs and the Particle Size
We calculated the gas/particle distribution coefficients for each particle size, K pi , using Eq. ( 6): where PM i is the concentration of particulate matter of the size of interest (µg m -3 ), and F i and A are the concentrations of the compound of interest in the particulate and gas phases, respectively.Plots of log K pi against log P L 0 for the PCDD/Fs and for the different particle sizes are shown in Fig. 5.The correlation coefficients tended to be higher for the finer particles than for the coarser particles.The regression slopes ranged from -1.152 to -0.897, and the intercepts ranged from -6.528 to -5.973.Similar slopes and intercepts were found in a previous study (Lee et al., 2008).The results showed that the PCDD/Fs were adsorbed onto the particles.In this study and in a previous study (Li et al., 2015a), the PBDFs were the main contributors to the PBDD/Fs.The PBDF distributions in the particles of different sizes were therefore also studied (Table 3).The correlation coefficients were higher for the particles of d ae 1.0-2.5 µm than for the coarse particles (d ae > 10 µm) and fine particles (d ae < 1.0 µm).The regression slopes ranged from -0.406 to -0.327, and the intercepts ranged from -3.541 to -3.192.The PBDFs were mainly absorbed into the particles in all of the particle size fractions, without any clear variations.

PCDD/F Homolog and PBDD/F Congener Profile Distributions for Different Particle Sizes
As mentioned above, the distributions of PCDD/Fs in particles of different sizes have been determined in many studies.For example, Kaupp et al. (1994) found very similar PCDD/F homolog distribution patterns for all particle size fractions in samples collected in a rural area.The lower chlorinated PCDD/F distribution has been found to be biased toward coarse particles and the higher chlorinated PCDD/F distribution toward fine particles in other studies (Kaupp and McLachlan, 2000;Oh et al., 2002).The low PCDD/F concentrations that are found in coarse particles may be a reason for the little information available.It is difficult to describe the general characteristics of the PCDD/F distribution in relation to particles of different sizes because few similar studies have been performed.
The PCDD/F homolog and PBDD/F congener distributions in particles of different sizes are presented in Fig. 6.The higher chlorinated PCDD/Fs (hexa-to octa-chlorinated) were mainly found in the particulate phase in all of the particle size ranges.The lower chlorinated PCDD/Fs (tetraand penta-chlorinated) fractions in the particulate phase increased as the particle size increased.The higher chlorinated PCDD/Fs were mainly found on the finer particles, consistent with the results of previous studies (Kurokawa et al., 1998;Oh et al., 2002).As shown in Fig. 6(b), the PBDFs, especially the higher brominated PBDFs, were the dominant PBDD/F congeners in the particles in every size range (Wang et al., 2008;Li et al., 2015a).The fraction of the higher brominated PBDFs in the particulate phase decreased as the particle size increased, but that was not the case for the PBDDs.More detailed studies on PBDD/Fs are currently in progress.

CONCLUSIONS
The PXDD/Fs were mainly found in the particulate phase in ambient air samples collected in a suburban part of Beijing.The fraction of the total PXDD/F concentration that was in the particulate phase increased as the particle size decreased.More than 80% of the total PXDD/F concentrations were in the particles of d ae < 2.5 µm.The regression coefficients for the relationship between K p and P L 0 for the PXDD/Fs indicated that the PCDD/Fs were mainly adsorbed onto the particles and the PBDD/Fs were mainly absorbed by the particles.The gas/particle partitioning parameters for the sum of the size fractions were similar to the parameters for particles of the specific size ranges that were collected.This implies that there were no specific sources of particles of certain sizes near the sampling site.If certain types of particulate matter are released into the environment it is possible that certain particle size fractions will vary in a particular way.Our results therefore emphasize the need for PXDD/Fs associated with particles of different sizes to be determined when assessing the environmental behavior and health effects of PXDD/Fs in the atmosphere.

Fig. 2 .
Fig. 2. Regression of log K P versus log P L 0 for the PCDD/Fs.
of the R 2 values were significant at P < 0.05 except for the value marked *, which was significant at P < 0.01.

Table 1 .
Meteorological information for the sampling period and the sample collection data.