A Q-filter was placed in a 10 mL glass flask with stopper and spiked with defined amounts of the internal standard (IS) substances. The IS substances employed in this study were as follows. For PAHs: anthracene-d₁₀ (Ant-d₁₀), pyrene-d₁₀ (Py-d₁₀), benzo(a)pyrene-d₁₂ (BaP-d₁₂), and benzo(g,h,i)perylene-d₁₂ (BghiP-d₁₂). For n-alkanes: n-tetracosane-d₅₀ (C₂₄-d₅₀). For PAEs: diethyl phthalate-d₄ (DEP-d₄), di-n-butyl phthalate-d₄ (DBP-d₄), bis(2-ethylhexyl)adipate-d₈ (DEHA-d₈), and benzyl butyl phthalate-d₄ (BBP-d₄). Samples were extracted in 5 mL dichloromethane solution twice using an ultrasonic bath for 15 min, followed by filtration through a hydrophilic PTFE syringe filter (ADVANTEC Co. Ltd., DISMIC, pore size of 0.22 μm) to remove insoluble particles and quartz fibers. The extracts were evaporated to dryness under a gentle stream of high purity nitrogen gas and re-dissolved in 100 μL of toluene for GC/MS analysis.

Analysis of the target organic compounds in the PM_2.5 samples was carried out by GC/MS (6890 GC/5973 MSD, Agilent Technology) under the following conditions: capillary column DB-5MS (length; 60 m, internal diameter; 0.25 mm, film thickness; 0.25 μm film thickness, Agilent J&W) was used for separation of the compounds; high-purity helium was used as the carrier gas with flow rate of 1.0 mL/min; sample injection amount was 1 μL in splitless mode; the GC oven temperature was initially held at 100°C for 1 min, increasing 20°C/min until 270°C and held for 17 min, increasing 15°C/min until 320°C and then held for 32 min; the injection temperature was set to 280°C and the MS source and quadruple temperature to 230°C and 150°C, respectively; and mass spectrum peaks were obtained by electron ionization mode at electron energy of 70 eV with a selected ion monitoring (SIM) mode. Monitored target and qualifier ions, and internal standards used in SIM mode is listed in Table 4.

The individual organic compounds and IS substances were determined by comparing retention times and mass spectrum peaks with those of authentic standards. The five-point calibration curves were plotted based on the IS method for individual compounds and all of the compounds showed good linearity with the Pearson’s correlation coefficients of R>0.95.

A recovery test (n=4) was performed to validate the pretreatment method. Known amounts of individual organic compounds were spiked to the blank Q-filter and then prepared and handled by exactly the same procedures as those used for the PM_2.5 samples. The average recovery for PAHs, n-alkanes and PAEs were 88-110%, 78-130% and 73-91%, respectively (Tables 2 and 3). In addition to the organic compounds described in Tables 2 and 3, we examined naphthalene (NaP), acenaphthene (Ace) and acenaphthylene (Acy) in PAHs, n-dodecane (C₁₂), n-tridecane (C₁₃), n-tetradecane (C₁₄), n-pentadecane (C₁₅), and n-hexadecane (C₁₆) in n-alkanes and diethyl phthalate (DEP), dibutyl phthalate (DBP) and dicyclohexyl phthalate (DCHP) in PAEs; however, the recovery percentages of those compounds were not acceptable. Therefore, these were eliminated from the analyte candidates in this study. The unacceptable recovery percentages for these compounds may have been due to vaporization loss during the evaporation process (NaP, Ace, Acy, and C₁₂-C₁₆) and a high level of background contaminations (DEP, DBP, and DCHP).

The limit of detection (LOD) for the individual organic compounds is provided in Tables 2 and 3. It was calculated using the standard deviation of standard solution values (n=7) multiplied by 3. Measurement data below the LOD were replaced with the value of LOD/2 for the calculations in this study.

PAHs are derived exclusively from the incomplete combustion and pyrolysis of organic materials contained in fossil fuels, coal fuels, lubricant oils, biomass, etc., and are ubiquitous environmental contaminants (Alves, 2008; Park et al., 2006; Simoneit, 2002). There are studies indicating or suggesting the linkage between PAH properties and human health problems such as cancer (IARC, 2013, 2010).

Measurement values of individual PAHs in each season are shown in Table 2. The summed concentration of 21 species of PAHs is expressed as Σ PAHs. Fig. 3 shows temporal variations of Σ PAHs obtained through the long time series analysis. Winter 2016 showed the highest average Σ PAH concentration (4.0 ng/m³) during the study period and other studies conducted in Japan support our results (Suzuki et al., 2015; Kume et al., 2007; Kakimoto et al., 2002). There were a few samples in early summer (July) and early winter (December-January) showing an elevated level of ΣPAHs; however, most of the samples were below 4 ng/m³ during the study period and the seasonal variation was little. Kume et al. (2007) have analyzed 21 species of PAHs in PM_2.5 from February 2001 to January 2002 in Shizuoka, Japan, and their result did not show strong seasonal variations, implying similarity to our result. This result suggests that the emission sources, which are not influenced by the seasons, are the dominant inputs for Σ PAHs within the source areas of the sampling site. The overall mean concentration of Σ PAHs was 2.9 ng/m³, which accounts for 0.030% and 0.19% of PM_2.5 and OC, respectively. Also, PM_2.5 and OC have positive correlation with Σ PAHs with the Pearson correlation coefficient (R) of 0.57 and 0.42, respectively. Our result is approximately one to two orders of magnitude lower than those observed in China (Wang et al., 2016, 2015). Although the individual amounts of organic compounds and their percentages accounting for PM_2.5 and OC are small, they could be used as molecular markers for source apportionment, which will be discussed later.

Fig. 3. 
Temporal variation of ΣPAHs concentration.

As is evident from Table 2, the lighter weight PAHs (Σ 3-rings=Fl+Phe+Ant) were much lower than that of the heavier weight PAHs (Σ 4-rings=Fluor+Py+BcP+BaA+Chry+7,12DaA; Σ 5-rings=BbF+BjF+BkF+BeP+BaP+3MC+DBahA; Σ 6-rings =IP+BghiP+DBalP+DBaiP+DBahP) throughout the year. This is probably because the 3-ring PAHs have relatively higher vapor pressures compared with the higher weight PAHs and tend to partition in the gas phase. The 4-ring PAHs can exist in both gas and particulate phases, whereas the 5-ring and 6-ring PAHs are mainly condensed/adsorbed onto particle phases (Zhang et al., 2016). Among the 4, 5, and 6-ring PAHs, 5-ring PAHs were found to be dominant throughout the year. In summer and autumn, average values of 6-ring PAHs were higher than those of 4-ring PAHs. However, the opposite results were obtained in spring and winter. These results could be attributed to the atmospheric temperature (summer>autumn>spring>winter, Table 2), since 4-ring PAHs have lower vapor pressures than 6-ring PAHs.

The highest average concentrations measured in each season among all the PAHs were as follows; BbF and IP in summer 2016, IP in autumn 2016, Fluor in winter 2016, Fluor in spring 2017, IP in summer 2017, and IP in autumn 2017. Although WHO has not drafted a guideline for BaP, it is classified as carcinogenic to humans (Group 1) by the IARC (2012), and is probably the most studied PAH. BaP concentrations were above the LOD for all the samples, but no BaP values exceeded 1 ng/m³ which is an annual mean value for the protection of human health as set out in the Directive 2004/107/EC (EU, 2004).

In an urban environment, n-alkanes are emitted from a large variety of sources, including the incomplete combustion of fossil fuels, lubricant oils, biomass burning, and biogenic sources such as wind erosion of the vascular plant waxes and direct suspension of pollen microorganisms (Yadav et al., 2013; Alves et al., 2009; Alves, 2008; Simoneit, 2002; Oros et al., 1999; Rogge et al., 1993a, 1993b, 1993c).

Measurement values of n-alkane homologues in each season are shown in Table 3. The sum of C₁₇-C₄₀ n-alkane concentrations is expressed as Σ n-alkanes. Fig. 4 describes the temporal variation of Σ n-alkanes during the study period and the relative abundances of individual n-alkanes normalized by C₂₉, which is one of the most abundant n-alkanes, are presented in Fig. 5 to visually confirm the distribution profile in each season. Winter 2016 showed the higher average Σ n-alkanes concentration (21 ng/m³) than the other seasons. A few samples in late autumn of 2016 (October-November) and early winter of 2016 (December-January) showed relatively elevated Σ n-alkanes concentration; however, most of the samples had concentrations below 20 ng/m³ (overall average of 14 ng/m³) during the study period and the seasonal variation was little. This result suggests that the emission sources, which are not influenced by the seasons, are the dominant inputs for Σ n-alkanes within the source areas of the sampling site. The highest Σ n-alkanes concentration was observed in late summer of 2017, when all of the n-alkane homologues appeared at high concentrations. The overall mean concentration of Σ n-alkanes accounted for 0.14% and 0.84% of PM_2.5 and OC, respectively. Also, PM_2.5 and OC have positive correlation with Σ n-alkanes with the Pearson correlation coefficient (R) of 0.57 and 0.80, respectively.

Fig. 4. 
Temporal variation of Σ n-alkanes concentration.

Fig. 5. 
Relative abundances of individual n-alkanes normalized by C₂₉ in each season.

PAEs are widely manufactured as plasticizers to make plastic materials flexible, and also in building materials, food, drinking water, soil, medical devices, cosmetics, children’s toys, and other products (Wang et al., 2017; Yao et al., 2016; Chou and Wright, 2006). Several epidemiological studies have demonstrated that human exposure to PAEs can induce adverse effects, and PAEs are suspected to be endocrine disruptors, which has stimulated public concern (Okamoto et al., 2011; Chou and Wright, 2006; Duty et al., 2005, 2003). To our knowledge, there is no previous publication involving the measurement of PAEs in PM_2.5 collected from the ambient atmosphere of Japan.

Measurement values of PAEs in each season are shown in Table 3. The summed concentration of 5 species of PAEs is expressed as Σ PAEs. Fig. 6 presents the temporal variation of Σ PAEs during the study period. Average Σ PAEs was found to be higher (0.51 ng/m³) in summer 2017 than in other seasons; however, average Σ PAEs in summer 2016 (0.26 ng/m³) was about half the summer 2017 value, so there was no clear seasonal pattern. One sample in late summer of 2017 showed a relatively high level of Σ PAEs (1.8 ng/m³); however, most of the obtained data were below 1 ng/m³ during the study period and thus no clear seasonal variation was observed. The reason is uncertain but the Σ PAEs were extremely low in autumn 2017. The overall mean concentration of Σ PAEs was 0.36 ng/m³, which was one and two orders of magnitude lower than Σ PAHs and Σ n-alkanes, respectively. Also, PM_2.5 and OC did not indicate correlation with Σ PAEs. Therefore, the target PAEs were not substantially emitted within the source area of the sampling site; however, PAEs those were not measured in this study might be contained in high concentrations in PM_2.5, so further research should target these nondetected compounds.

Fig. 6. 
Temporal variation of ΣPAEs concentration.

The diagnostic ratios of PAHs are a useful tool to identify possible sources and have been developed and applied by many environmental researchers (Wang et al., 2017, 2016; Hayakawa et al., 2016; Zhang et al., 2016; MikuŠka et al., 2015; Alves et al., 2009; Alves, 2008; Park et al., 2006; Brandenberger et al., 2005; Yunker et al., 2002; Simcik et al., 1999; Khalili et al., 1995; Li and Kamens, 1993; Rogge et al., 1993a, 1993b). The compounds involved in each ratio have the same or similar molecular weight, so it may be assumed that they have similar physicochemical properties. Diagnostic ratios vary during different stages of phase transfer and under environmental degradation. In this study, we calculated diagnostic ratios of BeP/(BeP+BaP), BaA/(BaA+Chr), IP/(IP+BghiP), Fluor/(Fluor+Py), and BghiP/BaP from the measured PAHs shown in Table 2.

Most fresh exhaust emissions typically contain almost the same amount of BeP and BaP. However, BaP more easily degraded than BeP by photochemical reactions in the ambient atmosphere. The atmospheric lifetime of BaP is estimated as ten times shorter than that of BeP (Kalberer et al., 2002). Therefore, the increment of BeP/(BeP+BaP) can be used as in indicator of the degree of photochemical degradation (i.e., aging) of compounds in the atmospheric environment (Wang et al., 2015; Shen et al., 2013; Alves et al., 2009). The seasonal variation of calculated BeP/(BeP+BaP) in this study ranged from 0.38-0.58. Freshly released BeP/(BeP+BaP) should have a ratio close to 0.50 (MikuŠka et al., 2015; Park et al., 2006), suggesting that most of the PM_2.5 containing BaP and BeP was emitted from local or nearby regional sources throughout the study period.

Table 5 presents five different types of developed diagnostic ratios from previous studies and includes BeP/(BeP+BaP), BaA/(BaA+Chr), IP/(IP+ BghiP), Fluor/(Fluor+Py), and BghiP/BaP (Hayakawa et al., 2016; Wang et al., 2015; Shen et al., 2013; Alves et al., 2009; Yunker et al., 2002; Simcik et al., 1999; Khalili et al., 1995; Li and Kamens, 1993; Rogge et al., 1993a, 1993b) and those obtained in this study. From the comparison of the diagnostic ratios shown in Table 5, most of the diagnostic ratios calculated were within the range of gasoline, diesel, coal and wood burning emissions. The emissions from gasoline and diesel emissions from vehicles, and coal emissions from industrial sources are not influenced by the seasons, which supports the assumption mentioned above in Section 3.2.1. Therefore, influences of vehicle and industrial emission sources nearby sampling site could be the possible anthropogenic sources of measured PAHs in this study. Kume et al. (2007) have analyzed 21 species of PAHs in suspended particles at urban sites of Shizuoka, Japan, within the area of vehicle and industrial emission sources nearby. Their annual means ratios of IP/(IP+BghiP) and BghiP/BaP were 0.51 and 1.28, respectively, which were in good agreement with those obtained in this study.

Moreover, careful consideration of PAHs usage data should be taken in to account when attempting to identify anthropogenic PAHs sources because PAHs can be emitted from various source types. Therefore, other molecular markers and/or tracers should be gathered for more accurate source apportionment.

Table 3 and Fig. 5 indicate that C₂₃-C₃₃ were the main homologues contributing to the Σ n-alkanes. Previous research has shown that carbon number with a strong predominance of odd carbon number indicates a significant contribution from biogenic contributors (Rogge et al., 1993c), whose abundances of odd carbon number higher than 24 are approximately one magnitude higher than those of even carbon number. Whereas n-alkanes released from fossil fuel sources do not show such predominant tendency (Alves et al., 2009; Simoneit et al., 1991). Fig. 5 shows that oddnumbered carbon homologues were not dominant during the study period. Therefore, contributions arising from anthropogenic inputs might have been higher than biogenic inputs at the study site.

For the further interpretation of obtained data associated with n-alkanes, we used the carbon preference index (CPI) and contribution of plant wax to n-alkanes (C_wax) to assess the variability of sources, and they are listed in Table 6. The CPI can be used to distinguish biogenic and anthropogenic inputs. It is defined as the sum of the concentrations of the odd carbon number n-alkanes (ΣC_odd) divided by the sum of the concentrations of the even carbon number n-alkanes (ΣC_even), which was calculated as follows:

Plant wax n-alkanes exhibit a strong odd carbon number predominance and thus increase the CPI value (Park et al., 2006). Conversely, anthropogenic inputs such as fossil fuels reduce the CPI value. The CPI typical values in the urban environment range from 1.1 to 2.0, while a CPI higher than 2.0 has a stronger biogenic influence (Alves et al., 2009). In the present study, the CPI values during the study period were in the range 0.73-3.2 with an average of 1.5, as shown in Table 6. A CPI value above 2.0 was observed just in one sample in spring 2017 (CPI=3.2, for 2-10 May), suggesting that most of the measured n-alkanes during the study period were anthropogenic in origin.

Another methodology to analyze the contribution of plant wax sources to n-alkanes can be expressed through C_wax (Wang et al., 2015; Yadav et al., 2013; Alves, 2008; Park et al., 2006; Simoneit et al., 1991). C_wax gives the percentage concentration input of wax n-alkanes from biogenic sources in sample. The C_wax is calculated as follows:

where n is the odd number of analyzed n-alkanes. Measured Cn values below the LOD were replaced with the value of LOD/2, and the negative values of {C_n -(C_n-1+C_n+1)/2} were determined as zero. Table 6 shows that the percentages of C_wax were in the range of 4.3-53% with an average of 22%. The higher percentage values indicate greater contributions from biogenic sources (Wang et al., 2015). The highest C_wax value of 53% was calculated from the same sample as CPI= 3.2, which could possibly mean that the contribution of biogenic sources was larger than sample during the period of 2-10 May 2017. Comprehensive evaluation including the historical distributions of n-alkanes homologues, CPI, and C_wax, indicates that the observed n-alkanes may largely originate from anthropogenic sources for most of the samples in this study. Further identification of the emission sources of n-alkanes in PM_2.5 will be conducted in future research.