Particulate Size Distribution and Sources Evaluation of n-Alkanes during Long-Term Haze Episode around Chaohu Lake , Eastern China

n-Alkanes (from nC16 to nC32) associated with particulate matters were determined in the ambient air around Chaohu Lake, Eastern China, from October to December in 2014 during a long-term haze episode. The total concentrations of particle bounded n-alkanes varied from 332 to 2500 ng m, with the homologues of nC24–nC30 the most abundant species. Spatial analysis revealed that low concentrations of n-alkanes existed at the sites close to Chaohu Lake, while high concentrations were generally at locations distant from the lake. For all aggregated fractions, most n-alkanes were distributed in fine particles with the mean geometric mean diameter (GMD) varying from 3.0 ± 0.6 μm for nC16 to 2.1 ± 0.6 μm for nC32. Short chain n-alkanes were accumulated in coarse particles with a unimodal distribution, but long chain aliphatic hydrocarbons appeared to have a biomodal distribution in fine and coarse particles. The mass size distribution of individual n-alkane homologue was predominantly influenced by its volatility; thus GMDs were well correlated with the logarithmically transformed subcooled liquid vapor pressures (PL, Pa) of n-alkanes at each sampling site, following the equation: GMDs = mgLogPL + bg. Furthermore, mg and bg obtained from all locations tended to exhibit a significant linear correlation. This suggests that all saturated aliphatic hydrocarbons follow a similar accumulation mode during a haze episode, which allows us to predict the size distribution and GMD of a compound based on its PL.


INTRODUCTION
With rapid economic growth and industrial development over the past three decades, China has experienced increasing environmental deterioration.Air pollution has become one of the most significant environmental and human health problems because overwhelming haze has frequently covered China's major cities in recent years and tens of millions people have suffered from high inhalation exposure to various air pollutants (Jiang et al., 2015;Li et al., 2015b).Atmospheric aerosol particles, especially those with an aerodynamic diameter less than 2.5 µm (PM 2.5 ), have been confirmed as vital air pollutants.These aerosols can not only be emitted from various anthropogenic sources including fossil fuel combustion, biomass burning, motor vehicles, but also formed from atmospheric reactions of organic volatile compounds or inorganic precursor gases (Yue et al., 2016' Cao et al., 2017).Natural mechanical processes including volcanic eruptions, wind-driven or traffic-related suspension of road, soil, and mineral dust, sea salt are also the important contributors to aerosol particles (Li et al., 2015a;Farahat et al., 2016).The atmospheric aerosols are generally composed of sulfate, nitrate, ammonium, sea salt, mineral dust, organic matter, and black or elemental carbon (Bzdek et al., 2012;Calvo et al., 2013).Organic matters such as aliphatic hydrocarbons are the important fraction of aerosol particles, usually accounting for 20-50% of total mass (Simoneit et al., 1991).
Aerosol particles can vary greatly in size (from less than 1 nm to 100 µm).The particles with different sizes are generally considered from different origins.Ultrafine particles (< 0.1 µm) can be emitted directly from various combustors or formed in the atmosphere by homogeneous nucleation.Fine particles (PM 2.5 ) can arise from combustion or formation from atmospheric reactions, whereas coarse particles are generally from the resuspension of dust or soil.Particle size is the key determining factor in the environmental behavior of aerosols and thus influences residence time, removal mechanisms, adsorption capacity and reaction activation.Both ultrafine and coarse particles are susceptible to be removed from atmosphere and have relatively short life-time.Ultrafine particles are easy to coagulate to form large sizes of particles and coarse particles can be removed by dry deposition during the transport, whereas fine particle can reside in atmosphere for a long period of time.As a result, the particles size distributions of aerosols from emission sources to background areas change gradually (Offenberg and Baker, 1999).Furthermore, due to large specific surface areas, fine particles have higher binding affinity to metals and semi-volatile organic compounds when compared with coarse particles (Offenberg and Baker, 2000;Karanasiou et al., 2007).Interaction and partitioning between the gas and particle phase also depend on particle size (Kadowaki, 1994;Fraser et al., 1997).For example, less volatile organic compounds such as polycyclic aromatic hydrocarbons are more likely associated with fine particles (Hien et al., 2007;Ladji et al., 2014).More importantly, particle size greatly influences the health impacts of aerosols because fine particles can penetrate deep into lungs and bloodstream.Thus the size of particles which is usually described by their aerodynamic diameter (geometric mean diameter, GMD) has a key influence on their properties, chemical composition, fate and environmental/health impacts.Research on the size of aerosol particles and their chemical compositions is critical to the understanding of their atmospheric properties and effects on health and to the elucidation of their sources and environmental sinks.Most available studies were conducted to investigate the particle size distribution of PAHs (Offenberg and Baker, 1999;Hien et al., 2007), n-alkanes (Karanasiou et al., 2007), inorganic elements (Balachandran et al., 2000) and water-soluble compounds (Bao et al., 2009) in urban and rural areas, but knowledge has remained limited in the size distribution of particle bounded n-alkanes in typical haze episodes.
Therefore, the main purpose of this work was to elucidate the distribution of saturated aliphatic hydrocarbons (n-alkanes) associated with size segregated atmospheric aerosols collected around Chaohu Lake, Eastern China, during a long-term haze pollution episode.Additionally, we aimed to investigate whether the volatility of individual target compound affects its mass size distribution.

Study Areas and Samples Collection
In the present study, eleven sampling locations around Chaohu Lake were selected to collect the airborne particles during a long-term haze episode from October to December in 2014.These sampling sites included Changfeng city (CF), Xiaomiao town (XM), Emerald Lake campus of Hefei University of Technology (GDXQ), Feidong city (FD), Hefei NO.1 High School (YZ), Sanhe town (SH), Zhongmiao town (ZM), Chaohu city (CH), Sanbing town (SB), Tangchi village (TC) and Lujiang city (LJ) (Table S1 and Fig. 1).Generally, CF, FD, CH, and LJ are located in the downtown areas of the cities; GDXQ and YZ in the suburban areas of Hefei City; XM, SH, ZM, and SB in the central areas of cities; and TC in a rural area.It should be noted that SH is a typical tourist area.Generally, these sampling sites were distributed evenly around Chaohu Lake, allowing us to characterize the distribution of n-alkanes around Chaohu Lake.Air samples were collected using an eight-stage nonviable impactor (Anderson, Franklin, Massachusetts, USA) for a period of 72 hours.The impactor was deployed on the top of the buildings with a height of 5-30 m (Table S1).Except for GDXQ and YZ where the samplers were on the top of university buildings, the remaining samplers were placed on the top of hotel buildings.Airborne particles were sampled at a flow rate of 28.3 L min -1 and the cutoff aerodynamic diameters for each stage were 0.4-0.7,0.7-1.1, 1. 1-2.1, 2.1-3.3, 3.3-4.7, 4.7-5.8, 5.8-9.0, and > 9.0 µm.
Before sampling, glass fiber filters (Whatman International, Maidstope, UK) were baked in a muffle at 400°C for 4 hours to remove residual organic matters and then weighed.The filters after sampling were weighed again and packed with aluminum foils and stored at -20°C until analysis.For n-alkanes analysis, fiber filters were cut into small pieces, spiked with nC 24 -d 50 as recovery surrogates, and then Soxhlet extracted with 100 mL of a mixture of n-hexane and dichloromethane (1:1, v/v) for 48 hours.The extract was concentrated to about 1 mL and cleaned through a glass column packed with cotton, 3 g of silica gel and 1 g of anhydrous sodium sulfate from bottom to top.The final eluant was concentrated to about 1 mL and spiked with a known amount of internal standard (nC 30 -d 62 ) prior to instrumental analysis.
The quantitation and quantification of the target compounds were performed with a Model 7890A gas chromatograph (GC) coupled with a Model 5975C mass spectrometer (MS) (Agilent Technologies, USA).Gas chromatographic separation was performed on a 30-m DB-5MS capillary column (30 m × 0.25 mm, internal diameter; 0.25 µm film thickness).The MS was operated with an electronic impact (EI) mode at 70 eV using selected-ion-monitoring (SIM).The MS ion source and quadrupole temperature were set as 300°C and 180°C respectively.The injection volume was 1 µL and the injection was performed in splitless mode at 300°C.The oven temperature was programmed from 60 to 180°C at a rate of 8 °C min -1 , to 220 °C at a rate of 2 °C min -1 , and then to 320°C at a rate of 8 °C min -1 .

Quality Assurance
All samples were spiked with surrogate standard to monitor the recovery of sample extraction.A method blank was included with every batch of 20 samples.The recoveries of surrogate standards ranged from 54.3 to 136%.Reported concentrations were not corrected with recovery efficiency.

Data Analysis
Carbon preference index (CPI), an important indicator of n-alkanes sources, is defined as the ratio of the total nalkanes with odd carbon number to those with even carbon numbers.It is usually used to differentiate the contribution of biogenic organics from anthropogenic emission.
In general, n-alkanes are considered to be from biological origins if a CPI value is greater than 3.0, whereas they are inferred from anthropogenic sources if a CPI value is close to unit (Simoneit, 1989).Apart from this, the contribution of plant wax deriving n-alkanes can be calculated with the following equation to determine the relative importance of biogenic or petrogenic sources (Wang et al., 2012): where %WaxC n (ng m -3 ) is the concentration of plant wax n-alkanes and negative values are taken as zero.C n , C n-1 and C n+1 are the concentrations of n-alkanes with neighboring carbon numbers.When the compositional characteristics of n-alkanes exhibit a shape of jaggedness, n-alkanes usually have a greater %WaxC n which suggests biological sources as the main source.If an odd to even preference doesn't exist and %WaxC n is smaller, it usually indicates the main source of anthropogenic emission.Geometric mean diameter (GMD) was applied to describe the size distribution of n-alkanes in this study and determined using the below equation: where C i is the concentration of n-alkanes in size class i and D p,i is the geometric mean particle diameter collected on stage i (Offenberg and Baker, 1999).

Mass Concentration of n-Alkanes around Chaohu Lake
According to the data released by the air quality monitoring stations of Hefei (http://www.pm25x.com/city/hefei.htm), the air quality index (AQI), PM 2.5 and PM 10 contents varied in ranges of 54.9-238, 37.3-187 ng m -3 and 49.5-296 ng m -3 , respectively (Table S1).The recorded highest daily aerosol concentrations were 187 µg m -3 for PM 2.5 and 296 µg m -3 for PM 10 , respectively, approximately twice the daily average standard of 65 µg m -3 for PM 2.5 and 150 µg m -3 for PM 10 regulated by the United States Environmental Protection Agency (U.S. EPA).
All n-alkane homologues from nC 16 to nC 32 were detectable in all samples collected around Chaohu Lake.Aggregate particle-associated n-alkanes (sum of all size fractions or ΣAK) varied from 476 to 2500 ng m -3 with an arithmetic mean of 1220 ± 812 ng m -3 .SB was observed with the lowest summed concentration, and samples collected in ZM, GDXQ, YZ and CH also contained relatively low levels, generally less than 1000 ng m -3 .However, the concentrations in SH, CF, XM, LJ, TC and FD were relatively higher with the maximum value in LJ, where the sampling site was located at a downtown area.Therefore, a general spatial distribution pattern was observed, i.e., low n-alkanes concentrations at the sites close to Chaohu Lake and high concentrations in the locations distant from the lake (Fig. 1).Relatively low n-alkanes concentrations in the sampling sites close to Chaohu Lake were partly associated with the different features of wind speed which was also important to the haze formation.Previous studies indicated that high particular matter concentrations were frequently observed in the meteorological condition with low wind speeds (Jiang et al., 2015;Li et al., 2015b).The average PM 2.5 concentrations recorded in Binhu Monitor Station (close to the YZ site) during the entire sampling period was significantly lower than that in Luyang Monitor Station (close to the CF site, p < 0.05) and the mean value in Gaoxin Monitor Station (close to the XM site, p < 0.05) (Fig. 2).Spatial variation in the summed n-alkanes concentrations could also be attributed to the open straw burning and residual coal combustion in the outlying regions of the city during the sampling period (Wang et al., 2009a;Wang et al., 2009b).

Compositional Characteristics of n-Alkanes around Chaohu Lake
The compositional distributions of n-alkanes for all aggregated fractions at eleven sites showed that homologues from nC 24 to nC 30 were overall dominant but their abundances were also site dependent (Fig. 3).The carbon number of the most abundant n-alkanes (C max ) was at nC 26 in LJ, XM, CF and FD, at nC 27 in SH and at nC 29 in the remaining locations.Interestingly, all sites with C max at nC 26 exhibited a characteristic unimodal n-alkane profile, possibly suggesting significant contributions from heavy duty diesel trucks (Pietrogrande et al., 2011).Indeed, these sampling sites were close to the major transits.By contrast, the sites with C max at nC 27 or nC 29 had a bimodal distribution of n-alkane with the first peak at nC 17 and the second peak at nC 27 or nC 29 , which probably indicated a mixture of sources combining fossil fuel and biomass combustion with weak biogenic and petrogenic signatures (Lin and Lee, 2004).For all sites, the CPI ranged from 1.3 ± 0.2 (site TC) to 1.8 ± 0.5 (site ZM), suggesting significant anthropogenic activities related sources along with minor contribution from vascular plants.The contributions of biogenic wax n-alkanes to the total n-alkanes concentrations based upon the estimation of %WaxC n ranged from 10.5 ± 3.5% (site GDXQ) to 19.5 ± 7.9% (site ZM).Generally, the sample with a relatively higher CPI value usually had a relatively higher %WaxC n value, resulting in a significantly positive correlation between these two indicators (Fig. 4).
Compared with other studies, it was clear that the contribution of biogenic sources to particle bounded n-alkanes around Chaohu Lake in winter was similar to those reported in other Chinese cities.For instance, the CPI values reported Tianjin (China) in winter ranged from 1.07 to 1.49 and the %WaxC n 7.68%-22.69%,which reflected main sources from anthropogenic emission and less significant contributions from higher plants (Wu et al., 2007).Even lower %WaxC n values (7 ± 4%-9 ± 5%) were reported in Shanghai (China), revealing even lower contributions from biogenic sources in such a major metropolitan area (Feng et al., 2006).

Size Distribution of n-Alkanes
In the present study, the particles collected on the impactor stages with a diameter less than 2.1 µm were referred to as fine particles, while the particle diameter greater than 2.1 µm represented coarse particles.On average, fine  particles associated n-alkanes accounted for 52.1 ± 12.5% of the total concentrations, with the minimal proportion of 46.2% at XM and the maximum proportion of 67.7% at YZ.These results indicated that most n-alkanes were distributed in fine particles consistently across sampling sites (Fig. 5).
Generally, the fraction with a size of 0.4-0.7 µm contributed the largest percentage of particulate ΣAK contents at ZM and FD.For the samples collected at SB, XM and YZ, most n-alkanes were found to be associated to the fraction of 0.7-1.1 µm, accounting for 20.4%, 20.5% and 50.7% of the total ΣAK contents, respectively.However, n-alkanes at TC, LJ, CF, SH, CH and GDXQ were greatly dominated by the fraction of 1.1-2.1 µm with the contribution to the total ΣAK contents of 29.9%, 50.1%, 21.6%, 35.3%, 48.1% and 53.3%, respectively (Fig. 6).Because of the large surface areas of fine particles, contaminants preferentially absorb onto the surface of fine particles by gas-solid distribution equilibrium (Chen et al., 2006;Tian et al., 2011).Previous research also indicated that particles with small sizes were generally enriched with n-alkanes and the concentrations of n-alkanes in the smallest particles were more than 10 times greater than that in the largest particles (Yassaa et al., 2001;Bi et al., 2005).Fig. 7 shows the Lundgren diagrams (dC/CdlogD p vsD p ) for all samples collected at 11 sites, which is a useful method for comparing the concentrations of coarse and fine particles (Zhu et al., 2014).The n-alkanes in all sampling sites with the exception of ZM and SH revealed a bimodal distribution.The fine mode was in the size of 0.7-1.1 µm, occasionally in the range of 0.4-0.7 µm (at SH, YZ and FD).The coarse mode predominantly corresponded to 5.9-9.0 µm.By contrast, n-alkanes at the samples from ZM and SH exhibited a unimodal distribution with the peaks at 3.3-5.8µm and 1.1-3.3µm, respectively.It should be noted that n-alkanes at YZ, GDXQ , SH and CF were mostly accumulated in fine particles, possibly suggesting fossil fuel combustion derived hydrocarbons (Karanasiou et al., 2007).For the samples from the remaining sites, natural sources can also play a role on the size distributions (Ladji et al., 2014).
Additionally, n-alkane species with different carbon chain length differed greatly in their associations with fine or coarse particles.Generally, short chain alkanes were accumulated in coarse particles with a unimodal distribution from nC 16 to nC 23 , whereas long chain hydrocarbons could be associated with both fine and coarse particles with a Fig. 6.Concentrations of n-alkanes in each size fraction for all sampling sites around the Chaohu Lake basin.biomodal distribution (Fig. 8).Semi-volatile organic compounds in urban aerosols usually exhibited a biomodal distribution because of primary and second sources, but those in aerosols from long-term transport exhibited a unimodal distribution due to dry deposition of coarse fractions or repartition between gaseous and particulate phases (Offenberg and Baker, 1999).Therefore, significant desorption and readsorption occurred for short chain n-alkanes.Long chain n-alkanes could be from both anthropogenic and natural sources (Bi et al., 2005).Previous studies indicated that hydrocarbons associated with biogenic sources were usually accumulated in coarse fractions (Alves et al., 2000;Bi et al., 2005).Thus high CPI and %WaxC n values in the coarse fractions (> 9.0 µm) (Fig. 9) at all sampling sites suggested significant contributions from natural sources.The lowest CPI and %WaxC n values were found in the fraction of 1.1-2.1 µm, likely implying the primary emission from fuel fossil combustion.Interestingly, the finest fraction (0.4-0.7 µm) was determined to include a signature of biogenic wax nalkanes.Indeed, size distribution of semi-volatile organic compounds is controlled by their emitting sources, volatilization/re-adsorption affinities, and meteorological conditions (Cheruiyot et al., 2015).Primary aerosols from fuel fossil and biomass combustion are generally among the smaller fine particles in the atmospheric environment (Offenberg and Baker, 1999).

Geometric Mean Diameters
GMDs varied from 1.1 µm for nC 25 at FD to 5.3 µm for nC 20 at XM. Generally, the samples collected at ZM, SB, TC, FD, and XM had larger GMDs than those from the remaining sites.Average GMDs for all samples ranged from 3.0 ± 0.6 µm for nC 16 to 2.1 ± 0.6 µm for nC 32 , tending to decrease along with the increasing of the number of carbon.GMDs in all sampling sites with exception of CH positively correlated against logarithmically transformed subcooled liquid vapor pressure (LogP L o ), whereas the GMD at CH was negatively correlated with LogP L o .Thus GMDs could be predicted by LogP L o with a general equation of GMDs = m g LogP L o + b g .The correlative regressions were significant (p < 0.05) at all sites and the values of m g and b g were summarized in Table 1.The value of m g varied from -0.053 to 0.254, which was significantly lower than those obtained for n-alkanes in urban aerosols (Offenberg and Baker, 1999).Aerosols in rural areas usually had m g much less than those in urban areas (Offenberg and Baker, 1999).Our results possibly suggested the aerosols from long-term transport during the haze episode.A significant correlation was found between m g and b g (Fig. 10), suggesting a general tendency of the correlation between GMD and LogP L o from a variety of samples collected around Chaohu Lake and the suitability of predicting GMD from P L o for a given organic compound.

CONCLUSIONS
Atmospheric particulate matters were collected in 11 locations around Chaohu Lake, Eastern China, from October to December in 2014 using an eight-stage non-viable impactor to determine 17 n-alkane homologues (from nC 16 to nC 32 ).Aggregated concentrations of particle-associated n-alkanes ranged from 332 to 2500 ng m -3 .The sampling sites with lower concentrations were close to Chaohu Lake and higher concentrations occurred at sites distant from the lake.These aliphatic hydrocarbons were predominately from anthropogenic activities along with minor contributions from biogenic source.The n-alkanes homologues, nC 24 to nC 30 , were dominant in the investigated samples from the study region.Particulate n-alkanes were mostly enriched in fine particles with the mean GMD ranging from 3.0 ± 0.6 µm for nC 16 to 2.1 ± 0.6 µm for nC 32 .Furthermore, our data demonstrated that LogP L o appeared to be an important factor influencing the size distributions of n-alkanes. .

Fig. 1 .
Fig.1.The spatial variation on the atmospheric particle bounded n-alkanes around Chaohu Lake.The area of circle represents the summed concentrations (ng m -3 ) of all n-alkanes homologues.samplingsite

Fig. 2 .Fig. 3 .
Fig. 2. Daily concentration of PM 2.5 between Oct to Dec of 2014 from three air quality monitoring stations.Daily concentrations were computed by hourly ambient observation data (http://www.pm25x.com/city/hefei.htm).Different letters within a column indicate significant difference (p < 0.05).B i n h u M o n i t o r S t a t i o n

Fig. 5 .
Fig. 5. Relative abundances of summed concentrations of nalkanes in fine and coarse particles from the Chaohu Lake basin.

Fig. 7 .
Fig. 7. Size distribution of n-alkanes for all sites around Chaohu Lake.

Fig. 8 .Fig. 9 .
Fig. 8.The distribution of individual n-alkanes homologue in fine and coarse particles from the Chaohu Lake basin.

Fig. 10 .
Fig. 10.The correlation between m g and b g in the aerosols collected from the Chaohu Lake basin.

Table 1 .
Slope (m g ) and intercept (b g ) for the linear regression between geometric mean diameter and log-based subcooled liquid vapor pressure (LogP L o ) (GMD = m g LogP L o + b g ).R 2 and p represent the square of regression coefficient and probability value, respectively.