Neutral Poly-/perfluoroalkyl Substances in Air and Snow from the Arctic

Levels of neutral poly-/perfluoroalkyl substances (nPFASs) in air and snow collected from Ny-Ålesund were measured and their air-snow exchange was determined to investigate whether they could re-volatilize into the atmosphere driven by means of air-snow exchange. The total concentration of 12 neutral PFASs ranged from 6.7 to 39 pg m−3 in air and from 330 to 690 pg L−1 in snow. A significant log-linear relationship was observed between the gas/particle partition coefficient and vapor pressure of the neutral PFASs. For fluorotelomer alcohol (FTOHs) and fluorotelomer acrylates (FTAs), the air-snow exchange fluxes were positive, indicating net evaporative from snow into air, while net deposition into snow was observed for perfluorooctane sulfonamidoethanols (Me/EtFOSEs) in winter and spring of 2012. The air-snow exchange was snow-phase controlled for FTOHs and FTAs, and controlled by the air-phase for FOSEs. Air-snow exchange may significantly interfere with atmospheric concentrations of neutral PFASs in the Arctic.


Results
Neutral PFASs in Arctic air. High-volume air sampling (gas and particle phases) was carried out from September 2011 to September 2012 on a platform for atmospheric observation at the German station located in Ny-Å lesund (78u559N, 11u569E) ( Figure 1, Table  S2). A summary of the data for neutral PFASs concentrations in the Arctic atmosphere is given in Figure 2 and Table S3. The total concentration (vapor plus particle phases) of the 12 neutral PFASs (SnPFASs) in the Arctic atmosphere ranged from 6.7 to 39 pg m 23 (mean: 17 pg m 23 ). Figure 3 shows the concentration range, mean and media of individual neutral PFASs in Ny-Å lesund. Gaseous PFASs were dominant in all air samples and accounted for 91% of SnPFASs. The fraction of each neutral PFAS compound to SnPFASs is believed to be related the specific vapor pressure of each PFAS and will be discussed later. The concentrations of PFASs were comparable with the data reported by Shoeib et al., who investigated the atmospheric levels of neutral PFASs in the North Atlantic and Canadian Archipelago 20 . As expected, the measured concentrations were considerably lower than those of urban and semi-urban areas 21,22 . The most abundant chemical in air is 852 FTOH, representing 61% of SnPFASs, and next three most abundant compounds were 652 (13%), 1052 (12%) and 1252 FTOH (4.8%) ( Table S3). The concentrations of 652, 852, 1052 and 1252 FTOH (SFTOHs) in air varied from 5.6 to 34 pg m 23 with a mean value of 14 pg m 23 (Figure 3), which is lower than those in Toronto (mean 79.5 pg m 23 ) 21 , Hamburg (from 32 to 204 pg m 23 ) 22 , and the North Sea (84 pg m 23 ) 23 . The concentrations of MeFBSA, MeFOSA and EtFOSA (SFOSAs) in air were higher than the levels of MeFBSE, MeFOSE and EtFOSE (SFOSEs), which is in agreement with the results reported by Wang et al. 24 .
The average ratio of 852 to 1052 to 652 to 1252 FTOH in air obtained in this study was 4.850.951.050.4. The ratio can be considered as an indicator of long range atmospheric transport due to the longer atmospheric residence time of FTOH 852 (80 d) than that of FTOH 652 (50 d) 15 . Higher ratios were measured in the Arctic air compared with urban/-semi-urban areas, such as 1.150.251.0 for 852 to 1052 to 652 FTOH in semi-urban location in Toronto, Canada 21 , and 1.850.651.0 in the southeast of Hamburg, Germany 22 , while the ratios were similar to the reported at remote sites, such as in the  Mount Bachelor Observatory 15 , the Canadian Arctic 6 and the North Sea 23 . The results confirmed that the FTOHs in Arctic air were from long range atmospheric transport process.
To investigate the relationship between the variation of gasparticle partitioning of PFASs in the atmosphere and vapor pressure, the linear relationships between logK SP and logpu L for all the 45 air samples were estimated. The physicochemical calculator SPARC v4.6 (October 2011 release w4.6.1691-s4.6.1687) was employed to calculate the values of pu L for individual PFASs at different sampling temperatures (Table S2). Considering the pK a values of 652 and 852, FTA were below the environmentally relevant pH range (#7) and they mainly exist as anions in aerosols as stated by Wang et al. 24 . The corrected logK SP (neutral form) of 652 and 852 FTA were considerably lower than those of FTOHs, FOSEs and FOSAs with similar vapor pressures, thus, they were excluded in the regression calculation performed in this study. The comprehensive regression results show that logK SP correlates well with logpu L ( Figure S1). Table S2 presents the slopes, intercepts, relationship coefficients (r 2 ) and significance levels (p) of all the 45 samples. The slopes and intercepts ranged from 20.63 to 20.15 (mean: 20.37) and from 22.57 to 21.31 (mean: 21.90), respectively. The results confirm that the gas-particle partitioning of neutral PFASs (e.g., FTOHs, FOSAs and FOSEs) followed the classical logK SP -logpu L relation for classic POPs 24 .
Neutral PFASs in Arctic snow. Snow sampling was conducted on glaciers around Ny-Å lesund during January and May 2012 ( Figure 1, Table S4). The concentrations of neutral PFASs in the Arctic snow samples are summarized in Table S3 and Figure 4. The range of the total concentration was from 334 to 692 pg L 21 with an average value of 523 pg L 21 . To our knowledge, there are few reports on PFAS concentrations in snow, thus it is hard to compare these values with those from other regions. Similar to the situation in air, 852 FTOH is the dominant species in the Arctic snow, accounting for 45% of the total concentration. SFTOHs in snow were between 218 and 507 pg L 21 (mean: 369 pg L 21 ). The composition of the 12 PFASs in snow was different to that in air. Besides 852 FTOH; 1052 FTOH, MeFOSE and 1252 FTOH were the three most abundant species, representing 18%, 11% and 6.5% of the SPFASs   in snow, respectively (Table S3). The difference between the compositions in air and snow might be caused by different degradation processes for neutral PFASs in air and snow, and different air-snow exchange potentials of individual chemicals 23 . The average ratio of 852 to 1052 to 652 to 1252 FTOH was 56.2522.551.058.1 in snow. However, to date, there has been few data such ratios in snow, thus it is hard to compare the variation between Arctic and urban areas. The ratios in snow are considerably higher than those in air, and could be explained by the higher volatilization potential of 652 FTOH and by different degradation processes of FTOHs in air and snow.
Air-snow exchange fluxes of PFASs in the Arctic. Figure 5 shows the calculated net fluxes between air and snow for the 12 neutral PFASs from September 2011 to September 2012, whereby negative fluxes represent net deposition into snow and positive values are net volatilization fluxes into the atmosphere. It should be noted that the dominant trends of air-snow exchange were different among the various neutral PFASs species (Table S5) To investigate the influence of the uncertainties in C s , C a and SSA on the exchange fluxes and direction, the uncertainties of C s , C a and SSA were assumed to be 100%, and the exchange fluxes were calculated based on the Monte Carlo method using Software Oracle Crystal Ball 11.1. The cumulative frequency of Monte Carlo analysis (10 000 times) for the 12 PFAS compounds are presented in Figure S2.
The results indicate that the exchange direction did not change with 95% certainty for all neutral PFASs except for 652 FTOH and EtFOSA.
The evident variability in air-snow exchange fluxes can be attributed to the varying K SA values of the different neutral PFAS species, which results in their potential of retaining in snow. K SA is a useful parameter expressing how much of the PFASs can be retained in snow under equilibrium conditions with air, and is determined by a combination of snow and compound specific properties. As stated by Hansen et al., snowmelt and the increased temperature would lead to a decrease of K SA 25 . Considering all collected snow samples in this study were fresh, the physical properties of snow can be considered as being constant, thus only the compound properties would influence the values of K SA . Daly and Wania found that volatilization from snow is not a significant for chemicals with logK OA values higher than 10 (or logK SA . 0.5) at 25uC 18 . Accordingly, PFASs with a logK SA . 0.5 would is unlikely to be volatile in snow, and as expected, FOSEs (logK SA . 0.5) have a negative exchange fluxes between air and snow (Table S6).

Discussion
Previous studies have confirmed that POPs in snow were mainly from the atmosphere, and that snow has a high efficiency in scavenging vapor and particle-bound POPs out of the atmosphere due to their large surface area and the low temperatures promoting surface sorption. The positive FTOHs and FTAs fluxes observed in this study show the dynamic nature of their air-snow exchange, i.e., re-volatilization from the snow occurs rapidly following their wet deposition, which implies their strong volatilization potential. Compared with FTOHs, FTAs and FOSEs, the net air-snow exchange fluxes for FOSAs varied between deposition and volatilization (Table S5), which implies that FOSAs in air and snow are in dynamic equilibrium.
In theory, under conditions of increasing temperature and snowmelt, snow would lose capacity to hold sparingly water-soluble chemicals due to decreases in volume, A snow , and values of logK SA 25 . As a result, chemicals accumulated in snow would have the potential for volatilization. Daly and Wania indicated that the air concentration of semi-volatile organic pollutants could be considerably influenced by snowmelt, and a high concentration would be observed during or after snowmelt, especially in high latitude areas 18 . A significant correlation in this study was observed between the ambient temperature and the atmospheric concentrations of FOSEs and FOSAs (r 2 5 0.70, p , 0.01) ( Figure S3). However, the concentrations of FTOHs and FTAs did not show a significant correlation with the ambient temperature. The differences in behavior of the different chemicals are believed to be due to differences in their volatilization potential in snow as discussed above. Once FTOHs and FTAs have being scavenged into snow, they have strong volatilization potential to air. On the other hand, net volatilization of FOSEs and FOSAs out of snow occurs only under conditions of increasing temperature or snowmelt (and hence lowers K SA ). As a result, air concentrations of FOSEs and FOSAs are significantly influenced by temperature. Hansen et al. found that the increase in atmospheric hexachlorocyclohexane isomers concentrations were more pronounced than those for fluorene, phenanthrene and PCB-28 toward the end of the winter in the Canadian High Arctic 25 . In general, chemical fluxes between snow and air are highly dynamic. Those chemicals having higher logK SA values can be efficiently retained in snow, which could lead to an increase in their concentration in the atmosphere during or after snowmelt.

Methods
Air and snow sampling. Air sampling was conducted on the German atmospheric observation platform located in Ny-Å lesund (78u559N, 11u569E) (Figure 1), Svalbard from September 27, 2011 to September 21, 2012, and a total of 45 air samples were collected. A glass fiber filter (GFF: diameter, 150 mm, pore size, 0.7 mm) and a selfpacked polyurethane foam (PUF)/XAD-2 cartridge (PUF: Q5.0 cm 3 2.5 cm; 35 g XAD-2, particle size: 0.3-1.0 mm) were simultaneously employed for collecting air samples (gas-and particle-phase) using a high-volume air sampler ( Figure S4). Each sample was collected at ,15 m 3 h 21 for 7 days to obtain a sample volume of ,2500 m 3 . After sampling, cartridges and GFFs were separately sealed in air tight aluminum bags and stored at 4uC and 220uC prior to extraction, respectively. Field blanks consisted of GFFs and PUF/XAD-2 cartridges that were taken to the field site and handled in the same manner as real samples. Detailed information on the sampling dates, air volume, total suspended particulate and the average temperature for each sample are listed in Table S2. Snow sampling was conducted around Ny-Å lesund from January to May 2012 (Table S4). The sampling depth was 0-5 cm and the samples were collected within 4 h after snowfall, thus the snow samples can be considered as fresh. Snow was collected using a pre-cleaned 40 L stainless steel barrel. After collection, the snow was allowed to melt within 24 h in the lab (at ,15uC). Around 5-8 L melt water was extracted with a polymer resin cartridge (40 g PAD-3 packed in a glass column) at a flow rate of ,300 mL min 21 , and was used for the determination of neutral PFAS. PAD-3 columns were stored at 4uC and the GFF filters were stored at 220uC.
Sample extraction. Prior to extraction, internal standards mixture (IS, 2.5 ng) ( 13 C 652, 852 and 1052 FTOH, MeFOSA D3, EtFOSA D5, MeFOSE D7 and EtFOSE D9) was spiked into a PUF/XAD-2 cartridge and GFFs, respectively. A modified Soxhlet apparatus (MX extractor) ( Figure S5), specially designed for the self-packed PUF/ XAD-2 and PAD-3 column, was employed to extract PUF/XAD-2 and GFF for 16 h using dichloromethane (DCM). Extracts were evaporated to 5 mL with hexane as a keeper and 3 g Na 2 SO 4 was added to remove residual water. Samples were further reduced to 200 mL under a gentle stream of nitrogen and spiked with 1 ng of 951 FTOH as an injection standard.
Instrumental analysis. The 12 neutral PFASs were analyzed using an Agilent 6890 gas chromatograph -5973 mass spectrometry that was equipped with a 60 m SUPELCO WAXH 10 column (60 m 3 0.25 mm 3 1.0 mm). Measurements were made using a selective ion monitoring mode with positive chemical ionization (PCI). Methane was used as a reagent gas for PCI and helium (at a flow rate of 1.3 mL min 21 ) was used as carrier gas. The detailed instrumental parameters are presented in SI. The full names and other information of the 12 neutral PFASs determined in the present study are summarized in Table S1.
Quality Control. Detailed quality control processes and breakthrough tests for air samples were described elsewhere 23,24 , Method detection limits (MDLs) for neutral PFASs in air and snow samples were derived from mean blank values plus three times the standard deviation (Table S1). All results were recovery corrected for neutral PFASs with corresponding internal standards.
Calculation method of air-snow exchange flux. The air-snow exchange fluxes for the PFASs were calculated based on the modified Whitman two-film resistance model, the processes and methods are described in detail in the supplementary information. Briefly, the air-snow exchange flux (F gas , pg (s m 2 ) 21 ) was calculated as 25 : where C s (pg m 23 ) and C a (pg m 23 ) are the concentrations of target compounds in snow and air, respectively, v is an exchange velocity (m s 21 ), K snow-air is the snow-air partition coefficient and can be expressed by the following equation 25 : where K SA is the snow interface-air partition coefficient, SSA is the specific surface area (m 2 kg 21 ), and r (kg m 23 ) is the density of water.
As stated above, all snow samples collected in this study were fresh, thus, the snowpack can be considered to be homogeneous. Eqs. (1) and (2) show that snow physical parameters are key factors influencing F gas . In this study, snow physical parameters were used and/or calculated based on the empirical relationship developed by Legagneux et al. 26 . The overall or summation solute hydrogen bond acidity (Sa 2 H ) and the overall or summation solute hydrogen bond basicity (Sb 2 H ) were calculated according to Lyakurwa et al. 27 , and the solute gas-hexadecane partition coefficient (logL 16