Lutz Ahrens,*, Tom Harner,*, Mahiba Shoeib, Douglas A. Lane, and Jennifer G. Murphy INTRODUCTION

Transcription

1 pubs.acs.org/est Improved Characterization of Gas Particle Partitioning for Per- and Polyfluoroalkyl Substances in the Atmosphere Using Annular Diffusion Denuder Samplers Lutz Ahrens,*, Tom Harner,*, Mahiba Shoeib, Douglas A. Lane, and Jennifer G. Murphy Science and Technology Branch, Environment Canada, Toronto, Ontario, Canada M3H 5T4 Department of Chemistry, University of Toronto, Toronto, Ontario, Canada M5S 3H6 *S Supporting Information ABSTRACT: Gas-phase perfluoroalkyl carboxylic acids (PFCAs) sorb strongly on filter material (i.e., GFF, QFF) used in conventional high volume air samplers, which results in an overestimation of the particle-phase concentration. In this study, we investigated an improved technique for measuring the gas particle partitioning of per- and polyfluoroalkyl substances (PFASs) using an annular diffusion denuder sampler. Samples were analyzed for 7 PFAS classes [i.e., PFCAs, perfluoroalkane sulfonic acids (PFSAs), fluorotelomer alcohols (FTOHs), fluorotelomer methacrylates (FTMACs), fluorotelomer acrylates (FTACs), perfluorooctane sulfonamides (FOSAs), and perfluorooctane sulfonamidoethanols (FOSEs)]. The measured particulate associated fraction (Φ ) using the diffusion denuder sampler generally followed the trend FTACs (0%) < FTOHs ( 8%) < FOSAs ( 21%) < PFSAs ( 29%) < FOSEs ( 66%), whereas the Φ of the C 8 C 18 PFCAs increased with carbon chain length, and ranged from 6% to 100%. The ionizability of some PFASs, when associated with particles, is an important consideration when calculating the gas particle partitioning coefficient as both ionic and neutral forms can be present in the particles. Here we differentiate between a gas particle partitioning coefficient for neutral species, K p, and one that accounts for both ionic and neutral species of a compound, K p. The measured K p for PFSAs and PFCAs was 4 5 log units higher compared to the interpolated K p for the neutral form only. The measured K p can be corrected (to apply to the neutral form only) with knowledge of the pk a of the chemical and the ph of the condensed medium ( wet particle or aqueous aerosol). The denuder-based sampling of PFASs has yielded a robust data set that demonstrates the importance of atmospheric ph and chemical pk a values in determining gas-particle partitioning of PFASs. INTRODUCTION Per- and polyfluoroalkyl substances (PFASs) are ubiquitous contaminants whose occurrence in the environment is of special concern because of their potential for environmental persistence, bioaccumulation, and toxicity. 1,2 They have been widely used since the 1950s as surfactants in textile coatings, paper treatments, electronics, and fire fighting foams. 3,4 PFASs can be released into the environment during production, usage, and disposal stages of their life-cycle. 3 5 Once released into the atmosphere, PFASs are subject to transport and transformation processes. 6 The gas particle partitioning of PFASs plays a key role in their global atmospheric distribution and long-range transport (LRT) potential. Previous studies have focused on the LRT of the more volatile and neutral PFASs [i.e., fluorotelomer alcohols (FTOHs), perfluorooctane sulfonamides (FOSAs), and perfluorooctane sulfonamidoethanols (FOSEs)], 7,8 whereas less is known about the atmospheric fate of acidic PFASs [i.e., perfluoroalkane sulfonic acids (PFSAs) and perfluoroalkyl carboxylic acids (PFCAs)]. PFSAs are considered strong acids (i.e., with a pk a < 1), although the issue of whether PFCAs are strong acids has been debated Under typical environmental ph conditions, PFSAs and PFCAs are mainly dissociated into their ionic forms, though in the gas-phase they would occur as a neutral species. The calculation of the gas particle partitioning coefficient (K p ) from measurements of ambient air relies on accurate determinations of the gas- and particle-phase concentrations. 12 However, it has been difficult to measure K p because of reported sampling artifacts for PFCAs due to gas-phase and aqueous-phase sorption to filters [i.e., glass fiber filters (GFFs) and quartz fiber filters (QFFs)] and particulate matter present on the filter when using conventional high volume sampling techniques. This artifact results in an overestimation of the particle-phase concentration (i.e., a positive or blow-on sampling artifact). 13 Conventional air samplers are also susceptible to potential blow-off sampling artifacts, whereby chemicals can volatilize from the filtered particles and then are collected on the sorbent downstream of the filter. 14 Thus, the Received: March 6, 2012 Revised: May 14, 2012 Accepted: May 18, 2012 Published: May 18, American Chemical Society 7199

2 measurement of the gas- and particle-phase using conventional high volume air samplers can lead to a bias in the determination of K p. The diffusion denuder sampler was developed to avoid these sampling artifacts by collecting the gas-phase first, followed by the particle-phase. 15 To our knowledge, there have been no studies investigating the gas particle partitioning for PFASs using diffusion denuder samplers. Another key consideration when calculating the gas particle partitioning of PFASs based on ambient air measurements is that some PFASs are ionizable which may lead to their enhanced partitioning and dissociation into ionic forms. This effect is expected to be greatest for wet particles and aqueous aerosols whereas partitioning to dry terrestrial particles is expected to be negligible The extent of partitioning and dissociation of PFASs will depend largely on their pk a values and the ph of the aqueous phase. The aim of this study is to determine and compare the gas particle partitioning of seven PFAS classes [i.e., PFSAs, PFCAs, FTOHs, fluorotelomer methacrylates (FTMACs), fluorotelomer acrylates (FTACs), FOSAs, and FOSEs] in the atmosphere on the basis of measurements using both diffusion denuder and conventional high volume air samplers. This will yield new insights into sampling artifacts associated with conventional air samplers and provide accurate data for assessing partitioning of ionizable PFASs. EXPERIMENTAL SECTION Chemicals. The target analytes included C 4,C 6,C 8,C 10 (PFBS, PFHxS, PFOS, PFDS) PFSAs (C n F 2n+1 SO 3 H), C 4 C 12, C 14 (PFBA, PFPeA, PFHxA, PFHpA, PFOA, PFNA, PFDA, PFUnDA, PFDoDA, PFTeDA) PFCAs (C n F 2n+1 COOH), 6:2, 8:2, 10:2 FTOHs (C n F 2n+1 CH 2 CH 2 OH), 6:2 FTMAC (C 6 F 13 CH 2 CH 2 OC(O)C(CH 3 ) CH 2 ), 8:2, 10:2 FTACs (C n F 2n+1 CH 2 CH 2 OC(O)CH CH 2 ), FOSA (C 8 F 17 SO 2 NH 2 ), methyl and ethyl FOSAs (C 8 F 17 SO 2 N(C n H 2n+1 )H), and methyl and ethyl FOSEs (C 8 F 17 SO 2 N(C n H 2n+1 )CH 2 CH 2 OH). In addition, 16 mass-labeled compounds were used as internal standards (IS) and N,N-dimethyl perfluorooctane sulfonamide (Me 2 FOSA, C 8 F 17 SO 2 N(CH 3 )(CH 3 )), 13 C 8 PFOS, and 13 C 8 PFOA were used as injection standards (InjS). Details are provided in Tables S1 and S2 of the Supporting Information (SI). Sampling. Air samples were collected using colocated annular diffusion denuder and high volume air samplers. The sampling site was a semiurban location in Toronto (Environment Canada field site, N, W) where 14 pairs of samples were collected over 24 h periods in November and December Annular diffusion denuder air samples were collected using an integrated organic gas and particle sampler (IOGAPS) system ( 120 m 3 over 24 h periods). 15,19,20 Compounds in the gas-phase were collected using two eight-channel annular denuders (dimensions, 52 mm diameter 285 mm length) in tandem followed by a filter pack for the collecting the particlephase. The surfaces of the denuder are coated with highly sorptive ground XAD-4 with a coated area of 3300 cm 2 for each annular denuder (Supleco, Bellefonte, PA). The filter pack consists of a GFF (type A/E Glass, 90 mm diameter, >1.0 μm particle retention, Pall Corporation, Quebec, QC, Canada) for collecting the particles followed by two sorbent impregnated filters (SIFs) in series using 0.36 mg XAD-4 per square centimeter QFF (90 mm diameter, Whatman, Piscataway, NJ). The SIFs have a higher retention capacity compared to GFF alone and capture analytes that may blow-off the GFF. As an additional precaution this is further backed-up by one more GFF to trap any other analytes that breakthrough the filter pack. The last GFF can also be used as a method blank. In parallel with the diffusion denuder sample collection, high volume air samples ( 340 m 3 over 24 h periods) were collected using a PS-1 type sampler ( 25 μm particle size cut; Tisch Environmental, Cleves, OH). The high volume air sampler uses GFFs (type A/E Glass, 102 mm diameter, >1.0 μm particle retention, Pall Corporation) for collecting the particle-phase followed by a polyurethane foam (PUF)/XAD-2 cartridge for trapping the gas-phase compounds. The PUF/XAD-2 cartridge consisted of 15 g of XAD-2 resin (Supelpak-2, precleaned from Supleco) sandwiched between a PUF plug (76 mm diameter and 60 mm thick, precleaned from Supelco) that was cut in half. Blanks for the denuder samples were collected by extracting the denuder tubes in the same manner as the real samples before deployment. GFF, SIF, and PUF/XAD-2 cartridge field blanks were collected by exposing them for 1 min at the sampling site and then treating them like real samples. All samples were stored at 20 C until extraction. Details of the sampling dates, air volume, meteorological data, and methodology for determining the total suspended particles (TSP) and their values are presented in Table S3 in the SI. Sample Extraction and Instrumental Analysis. Prior to extraction, the denuder samples, PUF/XAD-2 sandwiches, SIFs, and GFFs were spiked with 25 ng (mass-labeled FTOHs, FOSAs and FOSEs) and 5 ng (mass-labeled PFSAs and PFCAs) (absolute amount) of an IS mixture containing 16 mass-labeled PFASs (Table S2 in the SI). Each of the denuder samples was extracted separately twice with petroleum ether (2 150 ml) followed by two extractions using methanol (2 150 ml). The extraction was done by placing an end-cap on one end of the denuder, adding the solvent to the denuder, and then, after capping the other end, rolling them back and forth 10 times on an even surface to ensure that the solvent comes into contact with all of the annuli. PUF/XAD-2 sandwiches were Soxhlet extracted with petroleum ether/acetone (85/15, v/v) for 6 h, followed by a 16-h extraction with methanol. The SIFs, GFFs, and sqffs were extracted by sonication: twice with dichloromethane and then three times with methanol. The petroleum ether, petroleum ether/acetone, and dichloromethane extracts containing the more volatile PFASs (i.e., 6:2 FTMAC, FTACs, FTOHs, FOSAs, and FOSEs) and the methanol extracts containing the PFCAs and PFSAs were concentrated by rotary evaporation followed by gentle nitrogen blow-down to 0.5 and 1 ml, respectively, using iso-octane as the keeper solvent. Prior to injection, 10 ng of absolute Me 2 FOSA was added to the petroleum ether, petroleum ether/ acetone, and dichloromethane extracts, respectively, and 4 ng of absolute 13 C 8 PFOS and 13 C 8 PFOA were added to the 200 μl methanol extract in a polypropylene (PP) vial (Canadian Life Science, Peterborough, ON, Canada). The methanol extracts were further filtrated using Mini-Uniprep PP filters (0.2 μm pore size, Whatman, Piscataway, NJ) and finally transferred to PP vials. Additional details are given in Ahrens et al. 20 Analysis of 6:2 FTMAC, FTACs, FTOHs, FOSAs, and FOSEs was performed using gas chromatography (Agilent 7890A; Agilent Technologies, Palo Alto, CA) mass spectrometry (Agilent 5975C; Agilent Technologies, Palo Alto, CA) (GC MS) in selective ion monitoring (SIM) mode using 7200

3 Figure 1. Comparison of total PFAS concentrations (sum of gas- and particle-phases) measured by diffusion denuder and high volume air samplers (n = 14). Concentrations which differ by more than a factor of 2 between the two sampling techniques are indicated with an asterisk. Note logarithmic y-axis. positive chemical ionization (PCI). 20 Aliquots of 2 μl were injected on a DB-WAX column (30 m, 0.25 mm inner diameter, 0.25 μm film, J&W Scientific, Folsom, CA) (Table S4 in the SI). The separation and detection of the PFCAs, PFSAs, and PFOSA were performed by liquid chromatography (Agilent 1100; Agilent Technologies, Palo Alto, CA) using a triple quadrupole mass spectrometer interfaced with an electrospray ionization source in negative-ion mode (LC ( )ESI MS/MS; API 4000, Applied Biosystems/MDS SCIEX, Foster City, CA). 20 Aliquots of 25 μl were injected on a Luna C 8 (2) 100A column (50 mm 2 mm, 3 μm particle size; Phenomenex, Torrance, CA) using a gradient of 250 μl min 1 methanol and water [both with 10 mm aqueous ammonium acetate solution (NH 4 OAc)] (Table S5 in the SI). The isotope dilution method was used, which is based on the ratio of the peak areas of the target analyte to the IS. As the analytical standards were not available for C 13,C 15,C 16, and C 18 PFCA, quantification for these analytes was based on the MS/ MS parameters of C 14 PFCA (PFTeDA). Hence, the results given for these PFCAs should be considered as semiquantitative. The blank concentrations and limits of detection (LODs) are given in Tables S6 and Table S7 in the SI for the diffusion denuder sampler and for the high volume air sampler, respectively. Average recoveries were 81% and 66% for gasphase compounds using the diffusion denuder and high volume air samplers, respectively. For the particle-phase, average recoveries were 87% for either sampling technique (for details see Table S8 in the SI). A comprehensive validation of the above method is described elsewhere. 20 RESULTS AND DISCUSSION PFAS Concentrations in the Gas- and Particle-Phases. The PFAS concentrations in the gas- and particle-phases derived using the annular diffusion denuder and high volume air samplers are compared in Figure 1 (for details see Tables S9 and S10 in the SI). Overall, all of the 29 targeted PFASs were detected in air samples. However, the concentration and composition of the PFASs varied between the two sampling techniques (annular diffusion denuder vs high volume air sampler) as did the relative proportions and detection frequency of compounds in the gas- and particle-phase for 7201 the two techniques. For example, PFSAs and C 7 C 15 PFCAs were detected more frequently in the gas-phase using the diffusion denuder sampler while frequency of detection for the particle-phase was similar between the diffusion denuder and high volume air samplers (Table S11 in the SI). Furthermore, the diffusion denuder derived concentrations for PFSAs and PFCAs were 2.5 times higher in the gas-phase and 4 times lower in the particle-phase compared to results from the high volume air sampler. Generally, total air concentrations for PFASs measured using the diffusion denuder and high volume samplers were in good agreement (Figure 1). However, substantial differences were observed for measurements of PFAS gas particle partitioning using the two sampling approaches. The most abundant PFAS class in the gas-phase was the FTOHs for both sampling techniques, representing about 80 84% of the ΣPFASs. The next most abundant compound classes were PFCAs (11 16%) and FTACs ( %) (see Table S9 in the SI). Gas-phase concentrations for ΣFTOHs ranged from 39 to 153 pg m 3 and were in the same range as reported for suburban or urban areas, whereas the ΣFOSA and ΣFOSE concentrations ( pg m 3 and pg m 3, respectively) were about 1 order of magnitude lower than previously reported. 7,8,21 This reduction in air concentrations might be due to the phase-out of perfluorooctyl sulfonyl fluoride (POSF), reduced PFAS emissions by optimization of the production process, 4 or a production shift to shorter chain PFASs and new fluorinated chemicals. 22,23 It is interesting to note that the total ΣFTAC concentrations in this study were higher ( pg m 3 ) compared to the ΣFOSAs and ΣFOSEs. This demonstrates the importance of FTACs as the next most abundant precursor class after the FTOHs. Furthermore, PFOS was the dominant PFSA which is in agreement with recent measurements. 5,24 PFBA was the dominant compound for the PFCAs ( pg m 3 ), with a tendency of decreasing air concentrations for the longer chain PFCAs. The C 4 -based PFBA and PFBS are the main replacement compounds of the voluntary phase-out C 8 -based products (i.e., PFOA and PFOS) which may explain its elevated concentrations in air. 3,4 In the particle-phase (see Tables S9 and S10 in the SI), concentrations of the PFCAs, FTOHs, and FOSAs were more

4 Table 1. Measured Gas Particle Distribution and Predicted Acid Dissociation Constant (pk a ) of PFASs Using Annular Diffusion Denuder and High Volume Air Samplers diffusion denuder sampler high volume air sampler SPARC pk a a c total(neutral+ionized) c neutral c total(neutral+ionized) c neutral Φ b log K p (m 3 μg 1 ) c log K p (m 3 μg 1 ) d Φ b log K p (m 3 μg 1 ) c log K p (m 3 μg 1 ) d PFBS ± ± ± ± ± ± 0.49 PFHxS PFOS ± ± ± ± ± ± 0.24 PFDS ± 0 PFBA PFPeA PFHxA PFHpA PFOA ± ± ± ± ± ± 0.76 PFNA ± ± ± ± ± ± 0.59 PFDA ± ± ± ± 0 PFUnDA ± ± ± ± 0 PFDoDA ± ± ± ± 0 PFTrDA ± ± ± 0.37 PFTeDA ± ± ± ± 0 PFPeDA ± ± ± 0.18 PFHxDA ± 0 PFODA ± 0 6:2 FTOH ± ± ± :2 FTOH ± ± ± ± ± ± :2 FTOH ± ± ± ± ± ± :2 FTMAC :2 FTAC :2 FTAC FOSA ± 0 MeFOSA ± ± ± 0.38 EtFOSA ± ± ± ± ± ± 0.31 MeFOSE ± ± ± ± ± ± 0.22 EtFOSE ± ± ± 0.27 a pk a was calculated at 25 C using SPARC (September 2011 release w s ). b Φ = c p /(c g + c p ), where c p is the sum of the neutral and ionic PFASs in the particle-phase (pg m 3 ) and c g is the PFAS concentration in the gas-phase (pg m 3 ). c K p =(c p /TSP)/c g, where TSP is the total suspended particle concentration (μg m 3 of particles). d K p =(c p /TSP)/c g, where c p is the concentration of the neutral PFAS in the particle-phase (pg m 3 ). than 1 order of magnitude lower compared to the gas-phase, while FTACs and 6:2 FTMAC were not detected. The concentrations of ΣPFSAs and ΣFOSEs in the particle-phase were in the same range as for the gas-phase (i.e., <LOD 4.2 pg m 3 and pg m 3, respectively). ΣPFCA concentrations were generally low (<1.1 pg m 3 ) in the particle-phase, and short-chain PFCAs (<C 8 ) were not detected. Gas Particle Partitioning. Ionizable compounds like PFASs are expected to partition mainly into the aqueousphase when they are dissociated because of their hydrophilicity. 16,17 Consequently wet particles (particles with a water film) and aqueous aerosols may play an important role in the gas particle partitioning of these chemicals. The sampling in this study was performed under conditions of high relative humidity (RH, average values of 51 95%) which enhances the water content of aqueous aerosols. 25 Ionizable PFASs exist in the neutral form in the gas-phase whereas in aqueous aerosols and films they can dissociate to their ionic form. The degree of dissociation of the chemical depends on the ph of the condensing medium and the acid dissociation constant (pk a )of the compound. Gas particle partitioning is commonly quantified using the measured particle-phase concentration that accounts for both ionic and neutral species of a compound using the following expression K = p ( cp /TSP) cg (1) where K p is the measured gas particle partitioning coefficient, c p is the sum of the neutral and ionic PFASs in the particle-phase in ng m 3, c g is the gas-phase concentration of the neutral PFAS in ng m 3, and TSP is the mass concentration of particulate matter in μg m 3 (see Table S3 in the SI). Contrary to K p, a neutral species gas particle coefficient (K p ) can be defined as the partitioning between the gas- and particle-phases for the neutral form only. 12 To convert K p to K p, the measured c p needs to be corrected to derive the neutral fraction. This correction can be done with information on the ph of the aqueous layer and the pk a of individual PFASs using the following equation ph pka c = c /( ) p p where c p is the particle-phase concentration of the neutral PFAS in ng m 3. The pk a values for individual PFASs were calculated (2) 7202

5 using SPARC (September 2011 release w s ) (for details see Table 1 and Table S14 in the SI, which also compares SPARC estimations to previous work). 9 SPARC was used in this study for calculating the pk a values because only a few experimental data are available. 10,26 Furthermore, our pk a values derived using SPARC showed a good agreement with previously modeled data for all PFASs 9,27 29 except for the PFSAs 29 (for details see Table S14 in the SI). A ph of 4.5 was used on the basis of meteorological and speciation sampler data from the National Air Pollution Surveillance (NAPS) Network for a site operating continuously in downtown Toronto. The gas- and particle-phase data of major ions were input into an extended Aerosol Inorganics Model (AIM) 30,31 to estimate the ph in the Toronto atmosphere during November/December 2010 (additional details are provided in the SI text and in Table S15). After calculating the c p, the K p for the neutral form can be calculated as K p = ( cp/tsp) cg (3) A summary of the gas particle partitioning coefficients of individual PFASs is given in Table 1. The SPARC v4.6 estimated pk a values of the PFSAs, PFCAs, 6:2 FTMAC, and FTACs were all below 0.4, and therefore, they are mainly present as anions at a ph of 4.5. This explains why the measured K p values for PFSAs and PFCAs are so much higher (i.e., 4 5 log units) compared to the corrected K p (Table 1). It is important to note that, for PFASs having pk a values that would indicate they are neutral at an environmentally relevant ph range (i.e., FTOHs, FOSAs, and FOSEs), the corrected K p was not different than the measured K p (i.e., a correction was not required). The K p values for the FTOHs measured in this study were higher than previously reported at five different sites in Europe. 17,32 It is possible that the difference stems from dimerization or micellization of the molecule, or alternatively an experimental artifact in one of the two sampling methods that were used. These results illustrate the complexities and dynamic nature of gas particle partitioning of acidic PFASs. The extent of their partitioning to particles will be determined by the ph of the aqueous phase, in which the partitioning to the condensed state will increase with ph as there are more anionic species present at higher ph values (i.e., as long as ph exceeds pk a ). However, it should be noted that there are uncertainties associated with this approach as pk a and ph values typically apply to a pure aqueous phase whereas in the environment the aqueous phase will contain a mixture of organic and inorganic molecules which can influence dissociation and gas particle partitioning of PFASs. 18 Furthermore, the SPARC estimated pk a values may be prone to some error. The properties of particular PFAS classes and their extent of partitioning to particles depends on the functional group. For example, the log K p and log K p values of the FOSEs were log units higher than for the FOSAs with the same carbon chain length for both sampling techniques. The log K p and log K p for PFNA was 1 log unit lower than for PFOS using the diffusion denuder sampler, and the difference was more than 3 log units for the high volume air sampler (though this might also be related to sampling artifacts; see the Sampling Artifacts section). Furthermore, the perfluoroalkyl chain length had also a substantial influence on the gas particle partitioning of PFASs. The log K p for the PFSAs increased from 2.30 m 3 μg 1 (PFBS) to 1.42 m 3 μg 1 (PFOS) ( 6.7 to 5.78 m 3 μg 1 for the log K p ) and from 2.43 m 3 μg 1 (PFOA) to 1.03 m 3 μg 1 (PFPeDA) for the PFCAs ( 7.14 to 5.74 m 3 μg 1 for the log K p ) for the diffusion denuder sampler. Hence, with each additional CF 2 moiety the gas particle partition coefficient increased by 0.20 ± 0.08 log units for the PFCAs and PFSAs. It was reported that PFCAs can demerize in the aqueous phase. These PFCA dimers have a higher pk a than the monomer (and therefore also higher K p values), and it is expected that the dimerization concentration should decrease with increasing chain length. 11 This may explain why our measured K p values are higher than those previously estimated in models that do not account for dimerization 33 and may also explain the increasing K p with increasing chain length in our study. Sampling Artifacts. Blow-on sampling artifacts were reported for PFCAs and PFSAs using conventional high volume sampling techniques, 13,20 indicating substantial amounts of gas-phase compounds sorbed to the filter or trapped particulate matter instead of breaking through to the PUF sampler. This results in an overestimation of the particlephase concentration. Diffusion denuders overcome the blow-on artifact because the gas-phase is collected first followed by the particle-phase. Differences in the particulate associated fraction (Φ ) determined by using diffusion denuder and high volume air samplers should demonstrate the existence of sampling artifacts (as long as the blow-off errors are insignificant or accounted for, as they are in this case). The measured Φ is calculated using the following equation: cp Φ = cg + cp (4) The Φ of the FTOHs, FTACs, FOSAs, and FOSEs showed good agreement between the diffusion denuder and high volume air samplers (Figure 2). Generally, Φ followed the Figure 2. Comparison of average particle-associated fractions (Φ ) for PFASs measured by diffusion denuder and high volume air samplers. Averages for all samples are shown. The dashed line represents 1:1 agreement. trend FTMAC/FTACs (0%) < FTOHs (1 15%) < FOSAs (22 33%) < FOSEs (62 72%) for both sampling techniques. These ranges of particle-associated fractions for the various PFASs are in accordance with previous studies on PFASs in the atmosphere. 24,34 A key difference between observations in the results for the two techniques is that the high volume air sampler measurements resulted in much higher particle-associated fractions for PFBS, PFOS, and C 8 C 14 PFCAs compared to the diffusion denuder sampler (Figure 2 and Table 1). For example, PFOS had a particle-associated fraction of 100% for the high volume 7203

6 Figure 3. Particle-associated fraction (Φ ) versus the subcooled liquid vapor pressure (p L, Pa) of PFASs measured by annular diffusion denuder and high volume air samplers. Note the outlying FOSE data. samplers whereas only 46% was associated with particles using the diffusion denuder sampler. Similarly, C 8 C 14 PFCAs had a Φ of % using high volume air samplers versus only 6 61% for the diffusion denuder sampler (Table 1). These results highlight problems associated with using conventional high volume air samplers for assessing gas particle partitioning of some PFASs due to sorption of gas-phase chemical onto the GFF (i.e., the blow-on sampling artifact). 13 It should be noted that diffusion denuders are also subject to some sampling artifacts. First, sorption of particles to the denuder surface may occur and result in an overestimation of the gas-phase concentration (i.e., underestimated Φ values). The loss of particles due to sorption to the denuder walls was evaluated for the annular diffusion denuder used in this study. The results showed that particle losses due to sorption to the denuder would only occur for very small particles (i.e., <0.1 μm, see Figure S1 in the SI). During sampling, the potential for particle loss was further reduced by using a higher sample flow rate (i.e., 85 L min 1 ), which reduces the residence time of the particles in the denuder. The lack of a particle sampling artifact for the denuder is supported by the observed Φ for the FTOHs, FTACs, FOSAs, and FOSEs (these PFASs were not affected by blow-on sampling artifacts) which were similar for both sampling techniques (see Figure 2). Our measured Φ values are also in accordance with previous reported Φ for these PFASs in the atmosphere. 24,34 Second, gas-phase breakthrough tests were performed in a previous study which evaluated the performance of diffusion denuder samplers for PFASs showing that no substantial breakthrough exists for the denuder. 20 Additional information on denuder samplers is provided elsewhere. 15,20 Overall, the diffusion denuder sampler appears to be more robust and less subject to sampling artifacts for PFSAs and PFCAs compared to conventional high volume air sampler. Correlation of the Gas Particle Partitioning with the Subcooled Liquid Vapor Pressure (p L ), Temperature, and Relative Humidity (RH). Figure 3 shows the correlation of the measured Φ versus the log p L for individual PFASs at the average temperature during sample collection. The p L was calculated for individual PFASs using the physicochemical calculator SPARC v4.5 (version from September 2009), and overall, calculated values agree well with literature values (for details see Tables S12 and S13 in the SI). 33 It is interesting to note that the different versions of SPARC generate different estimates for the p L suggesting that the quality of the predicted data varies from version to version. For the diffusion denuder sampler, the measured Φ values for PFASs in Figure 3 (left panel) fit well and generate the typical sigmoidal curve, with the exception of the FOSEs. In contrast, the Φ values derived from the high volume air sampler did not show a clear sigmoidal pattern and were grouped according to Φ values of and This behavior is likely due to the sampling artifacts associated with high volume samplers. Although the correlations of Φ versus the log p L from this study generate the sigmoidal curves that have been observed for other semivolatile organic compounds (SVOCs), the curves are offset to the right by about 4 orders of magnitude higher p L compared to results for typical semivolatile, nonpolar compounds. 12 This apparent enhancement in partitioning to particles/aerosols is attributed to the fact that many of the PFASs are polar/ionizable chemicals that partition strongly to aqueous phases. Considering the gas particle partitioning models for polar/ionizable chemicals from Arp and coworkers, 17,18,32,35 a need exists to further develop empirical gas particle partitioning models for polar/ionizable chemicals. This topic is beyond the scope of the current study but will be pursued further in a follow-up study. Having established the superiority of the denuder derived gas particle partitioning data, further correlations of the gas particle partitioning (K p ) of the PFASs against p L were performed. The regression slopes for the PFASs ranged between 0.78 and 1.27, whereas the regression for the FOSEs was less informative and did not differ significantly from zero (p > 0.5, Pearson Correlation) (for details see Table S16 and Figure S2 in the SI). The scattering of the data in the correlation of the K p against p L may can be reduced by developing an aerosol sorption model which considers the sorption behavior of individual PFASs to the components of the condensed phase (i.e., particles and water). Certain PFASs may sorb preferentially to the aqueous phase while others to the nonaqueous phase. 18 The ph of aqueous aerosols in the atmosphere will play a key role in the gas particle partitioning of PFASs. However, the ph of aqueous aerosols can vary seasonally and depends on emissions sources to air. Khlystov et al. have shown a wide range of ph values for Pittsburgh and an importance of aqueous aerosol parameters such as RH. The ph of aqueous aerosols was shown to be more acidic under low RH conditions (RH < 40%). 25 Typically, the ph in aqueous aerosols for urban areas is lower than the value of 4.5 that was measured for Toronto during November/December Figure 4 shows that the K p for PFSAs and PFCAs can be explained by accounting for ionization in an aerosol of ph = 4.5. At lower ph values (i.e., <4.5) the regression line for the K p of the 7204

7 Figure 4. Log log plots of the K p versus the subcooled liquid vapor pressure (p L, Pa) for different PFAS classes. The K p value depends on the ph in the aqueous aerosol. This is an illustrative plot to show the dependency of the neutral species (that is available for transport in air) on the ph of the aerosol and the pk a. Estimations of K p (eq 3) were made using a range of ph values (eq 2) that are typical for urban air. A ph of 4.5 was based on aerosol measurements in Toronto (see SI text and Table S15). We note that there are currently large uncertainties with the pk a values that are used to estimate K p (eq 2) and a need for improvement in these determinations so that reliable models can be developed. PFSAs and PFCAs would shift upward (see Figure 4). These results demonstrate that K p for ionizable PFASs (e.g., PFSAs and PFCAs) is very sensitive to ph. 18 In contrast to this dynamic nature of the ionizable PFASs, the corrected K p for the neutral FTOHs, FOSAs, and FOSEs was not different than the measured K p for typical atmospheric ph values of about 0 7. Previous studies have shown that other parameters like RH and temperature can have an influence on the gas particle partitioning. 18 In this study, no significant relationship was found between the K p of individual PFASs and temperature (p > 0.05, Pearson Correlation). The lack of a correlation may be partly due to the relative small changes in temperature during the sampling period with average ambient air temperatures ranging from 9.4 to +7.7 C. However, there was a significant positive correlation of the K p with the RH for some PFASs (i.e., EtFOSA, C 10, C 12, and C 15 PFCA) (p < 0.01, Pearson Correlation). This indicates that the sorption of PFASs increases with higher RH. This is in agreement with a laboratory study showing that at a RH of 90% the sorption of polar compounds increases. 17 This is a further motivation for testing existing models 17,32,35 or developing new partitioning models for polar/ionizable chemicals that incorporate sorption to aqueous aerosols. Implications. The diffusion denuder sampler has been demonstrated as a more accurate method (compared to conventional high volume air samplers) for measuring the gas particle partitioning of PFSAs and PFCAs. It has the advantage of overcoming the blow-on sampling artifact (i.e., positive sampling artifact) by collecting the gas-phase first, followed by the particle-phase. It is important to note that PFASs can dissociate to their ionic forms when sorbed in aqueous films on wet particles or in aqueous aerosols. The extent of dissociation depends on the ph of the aqueous phase (i.e., water) and the pk a of the compound. As a result, the measured K p for the ionizable PFASs that have low pk a values (i.e., PFSAs, PFCAs, 6:2 FTMAC, and FTACs) were 4 5 log units higher than the K p values calculated for the neutral form The measured K p and calculated K p (for the neutral form) do not agree with the classical Junge Pankow model which is widely used for estimating the gas particle partitioning for nonpolar, nonionizable SVOCs. 12,36 Further studies are necessary to assess the adsorption and/or absorption behavior of PFASs to the particle surfaces and liquid films and the bulk phase of aqueous aerosols. The creation of mixed micelles for PFASs and the presence of other surface active agents in the aqueous aerosols will also impact the partitioning mechanism. 18 There is a need to use robust data sets of gas particle partitioning for polar/ionizable chemicals such as the PFASs to verify existing models 17,18,32,35 and to develop new models. Such models are required to evaluate the atmospheric longrange transport, deposition, and overall fate of PFASs in the environment. ASSOCIATED CONTENT *S Supporting Information Additional details on sampling sites, meteorological data, QA/ QC data, ph in aqueous aerosols, PFAS concentrations, and their modeled pk a and p L. This material is available free of charge via the Internet at AUTHOR INFORMATION Corresponding Author *Phone: (L.A.); (T.H.). Fax: (L.A.); (T.H.). lutz. ahrens@slu.se (L.A.); tom.harner@ec.gc.ca (T.H.). Notes The authors declare no competing financial interest. ACKNOWLEDGMENTS We thank Dr. Hans Peter H. Arp and Dr. Elisabeth Galarneau for discussion and helpful suggestions. We also thank Dr. Eric J. Reiner from Ontario Ministry of the Environment for support in the analysis of PFASs. Data used to calculate the ph of Toronto aerosols was obtained from the NAPS Network. Partial funds for this work were provided through the Chemicals Management Plan (Government of Canada), the Chemicals Management Division (Environment Canada), and from the UNEP Secretariat for the Stockholm Convention on POPs. REFERENCES (1) Martin, J. W.; Mabury, S. A.; Solomon, K. R.; Muir, D. C. G. Bioconcentration and tissue distribution of perfluorinated acids in rainbow trout (Oncorhynchus mykiss). Environ. Toxicol. Chem. 2003, 22, (2) Giesy, J. P.; Naile, J. E.; Khim, J. S.; Jones, K. C.; Newsted, J. L. Aquatic toxicology of perfluorinated chemicals. Rev. Environ. Contam. Toxicol. 2010, 202, (3) Prevedouros, K.; Cousins, I. T.; Buck, R. C.; Korzeniowski, S. H. Sources, fate and transport of perfluorocarboxylates. Environ. Sci. Technol. 2006, 40, (4) Paul, A. G.; Jones, K. C.; Sweetman, A. J. A first global production, emission, and environmental inventory for perfluorooctane sulfonate. Environ. Sci. Technol. 2009, 43, (5) Ahrens, L.; Shoeib, M.; Harner, T.; Lee, S. C.; Guo, R.; Reiner, E. J. Wastewater treatment plant and landfills as sources of polyfluoroalkyl compounds to the atmosphere. Environ. Sci. Technol. 2011, 45, (6) Ellis, D. A.; Martin, J. W.; Mabury, S. A.; De Silva, A. O.; Hurley, M. D.; Sulbaek Anderson, M. D.; Wallington, T. J. Degradation of

8 fluorotelomer alcohols: A likely atmospheric source of perfluorinated carboxylic acids. Environ. Sci. Technol. 2004, 38, (7) Shoeib, M.; Harner, T.; Vlahos, P. Perfluorinated chemicals in the Arctic atmosphere. Environ. Sci. Technol. 2006, 40, (8) Jahnke, A.; Ahrens, L.; Ebinghaus, R.; Temme, C. Urban versus remote air concentrations of fluorotelomer alcohols and other polyfluorinated alkyl substances in Germany. Environ. Sci. Technol. 2007, 41, (9) Goss, K. U. The pk a values of PFOA and other highly fluorinated carboxylic acids. Environ. Sci. Technol. 2008, 42, (10) Burns, D. C.; Ellis, D. A.; Li, H.; McMurdo, C. J.; Webster, E. Experimental pk a determination for perfluorooctanoic acid (PFOA) and the potential impact of pk a concentration dependence on laboratory-measured partitioning phenomena and environmental modeling. Environ. Sci. Technol. 2008, 42, (11) Cheng, J.; Psillakis, E.; Hoffmann, M. R.; Colussi, A. J. Acid dissociation versus molecular association of perfluoroalkyl oxoacids: Environmental implications. J. Phys. Chem. A 2009, 113, (12) Pankow, J. F. Review and comparative analysis of the theories on partitioning between the gas and aerosol particulate phases in the atmosphere. Atmos. Environ. 1987, 21, (13) Arp, H. P. H.; Goss, K. U. Irreversible sorption of trace concentrations of perfluorocarboxylic acids to fiber filters used for air sampling. Atmos. Environ. 2008, 42, (14) Galarneau, E.; Bidleman, T. F. Modelling the temperatureinduced blow-off and blow-on artefacts in filter-sorbent measurements of semivolatile substances. Atmos. Environ. 2006, 40, (15) Lane, D. A.; Johnson, N. D.; Barton, S. C.; Thomas, G. H. S.; Schroeder, W. H. Development and evaluation of a novel gas and particle sampler for semivolatile chlorinated organic compounds in ambient air. Environ. Sci. Technol. 1988, 22, (16) Pankow, J. F. Gas/particle partitioning of neutral and ionizing compounds to single and multi-phase aerosol particles. 1. Unified modeling framework. Atmos. Environ. 2003, 37, (17) Arp, H. P. H.; Schwarzenbach, R. P.; Goss, K. U. Ambient gas/ particle partitioning. 1. Sorption mechanisms of apolar, polar, and ionizable organic compounds. Environ. Sci. Technol. 2008, 42, (18) Arp, H. P. H.; Goss, K.-U. Gas/particle partitioning behavior of perfluorocarboxylic acids with terrestrial aerosols. Environ. Sci. Technol. 2009, 43, (19) Gundel, L. A.; Lane, D. A. Sorbent-coated diffusion denuders for direct measurement of gas/particle partitioning by semi-volatile organic compounds. In Gas and Particle Phase Measurements of Atmospheric Organic Compounds; Lane, D. A., Ed.; Gordon and Breach Science Publishers: Canada 1999; Chapter 11. (20) Ahrens, L.; Shoeib, M.; Harner, T.; Lane, D. A.; Guo, R.; Reiner, E. J. Comparison of annular diffusion denuder and high volume air samplers for measuring per- and polyfluoroalkyl substances in the atmosphere. Anal. Chem. 2011, 83, (21) Stock, N. L.; Lau, F. K.; Ellis, D. A.; Martin, J. W.; Muir, D. C. G.; Mabury, S. A. Polyfluorinated telomer alcohols and sulfonamides in the North American troposphere. Environ. Sci. Technol. 2004, 38, (22) D eon, J.; Crozier, P. W.; Furdui, V. I.; Reiner, E. J.; Libelo, E. L.; Mabury, S. A. Perfluorinated phosphonic acids in Canadian surface waters and wastewater treatment plant effluent: Discovery of a new class of perfluorinated acids. Environ. Toxicol. Chem. 2009, 28, (23) Mo ller, A.; Ahrens, L.; Sturm, R.; Westerveld, J.; van der Wielen, F.; Ebinghaus, R.; de Voogt, P. Distribution and sources of polyfluoroalkyl substances (PFAS) in the River Rhine watershed. Environ. Pollut. 2010, 158, (24) Vierke, L.; Ahrens, L.; Shoeib, M.; Reiner, E. J.; Guo, R.; Palm, W.-U.; Ebinghaus, R.; Harner, T. Air concentrations and particle-gas partitioning of polyfluoroalkyl compounds at a wastewater treatment plant. Environ. Chem. 2011, 8, (25) Khlystov, A.; Stanier, C. O.; Takahama, S.; Pandis, S. N. Water content of ambient aerosol during the Pittsburgh air quality study. J. Geophys. Res. 2005, 110, (26) Moroi, Y.; Yano, H.; Shibata, O.; Yonemitsu, T. Determination of acidity constants of perfluoroalkanoic acids. Bull. Chem. Soc. Jpn. 2001, 74, (27) Rayne, S.; Forest, K.; Friesen, K. J. Computational approaches may underestimate pka values of longer-chain perfluorinated carboxylic acids: Implications for assessing environmental and biological effects. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng. 2009, 44, (28) Steinle-Darling, E.; Reinhard, M. Nanofiltration for trace organic contaminant removal: Structure, solution, and membrane fouling effects on the rejection of perfluorochemicals. Environ. Sci. Technol. 2008, 42, (29) Wang, Z.; MacLeod, M.; Cousins, I. T.; Scheringer, M.; Hungerbuḧler, K. Using COSMOtherm to predict physicochemical properties of poly- and perfluorinated alkyl substances (PFASs). Environ. Chem. 2011, 8, (30) Clegg, S. L.; Brimblecombe, P.; Wexler, A. S. A thermodynamic model of the system H + -NH + 4 -SO 2 4 -NO 3 -H 2 O at tropospheric temperatures. J. Phys. Chem. A 1998, 102, (31) Wexler, A. S.; Clegg, S. L. Atmospheric aerosol models for systems including the ions H +,NH + 4,Na +,SO 2 4,NO 3,Cl,Br and H 2 O. J. Geophys. Res., [Atmos.] 2002, 107 (No. D14), art. no (32) Arp, H. P. H.; Schwarzenbach, R. P.; Goss, K. U. Ambient gas/ particle partitioning. 2: The influence of particle source and temperature on sorption to dry terrestrial aerosols. Environ. Sci. Technol. 2008, 42, (33) Arp, H. P. H.; Niederer, C.; Goss, K.-U. Predicting the partitioning behavior of various highly fluorinated compounds. Environ. Sci. Technol. 2006, 40, (34) Shoeib, M.; Vlahos, P.; Harner, T.; Peters, A.; Graustein, M.; Narayan, J. Survey of polyfluorinated chemicals (PFCs) in the atmosphere over the northeast Atlantic Ocean. Atmos. Environ. 2010, 44, (35) Arp, H. P. H.; Goss, K. U. Ambient gas/particle partitioning. 3. estimating partition coefficients of apolar, polar, and ionizable organic compounds by their molecular structure. Environ. Sci. Technol. 2009, 43, (36) Junge, C. E. Basic considerations about trace constituents in the atmosphere as related to the fate of global pollutants. In Fate of Pollutants in the Air and Water Environments;Suffet, I. H., Ed.; J. Wiley: New York, 1977; Part I, pp NOTE ADDED AFTER ASAP PUBLICATION Equations 1 and 3, and the footnotes of Table 1 were modified in the version of this paper published June 15, The correct version published June 20, 2012.