Particle

Environ. Sci. Technol. 2002, 36, 5218-5228

Study of the Effects of Particle-Phase Carbon on the Gas/Particle Partitioning of Semivolatile Organic Compounds in the Atmosphere Using Controlled Field Experiments BRIAN T. MADER† AND JAMES F. PANKOW* Department of Environmental Science and Engineering, OGI School of Science & Engineering, Oregon Health & Science University, P.O. Box 91000, Portland, Oregon 97291-1000

A controlled field experiment (CFE) methodology with a filter/sorbent sampler was used to minimize artifact effects when measuring values of the gas/particle (G/P) partitioning constant (Kp, m3 µg-1) for semivolatile organic compounds (SOCs) in the atmosphere. CFE sampling was conducted at three different locations (Beaverton, OR; Denver, CO; and Hills, IA). Kp values were measured for a series of polycyclic aromatic hydrocarbons (PAHs) and polychlorinated dibenzodioxins and dibenzofurans (PCDD/Fs). To examine the possible effects on the G/P partitioning of the amounts of organic material (om) phase, organic carbon (OC), and elemental carbon (EC) in the sampled particulate material, the measured Kp values were normalized by the aerosol mass fractions fom, fOC, and fEC according to Kp/ fom, Kp/fOC, and Kp/fEC. Using a log-log format, the resulting normalized values were all found to be more highly correlated with the subcooled liquid vapor pressure p°L than were the unnormalized Kp values. For the PAHs, the oneparameter model assuming Kp ) Kp,OC fOC yielded only slightly less variability in the predicted Kp values than did the one-parameter model Kp ) Kp,EC fEC. The twoparameter model Kp ) Kp,OC fOC + Kp,EC fEC was found to provide only small improvements over each of the oneparameter models. Overall, the data are more consistent with an absorptive mechanism of partitioning to the particulate material but do not rule out some role for adsorption to particle surfaces. The data suggest that small amounts of organic carbon (fOC ∼ 0.02) can have significant effects on the G/P partitioning of SOCs.

Introduction Gas/particle (G/P) partitioning in the atmosphere has been successfully parametrized using the partitioning coefficient Kp (m3 µg-1):

Kp )

cp cg

(1)

* Corresponding author phone: (503)748-1080; fax: (503)748-1273; e-mail: [email protected]. † Present address: 3M Environmental Technology and Safety Services, St. Paul, MN 55133-3331. 5218

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002

wherein cp (ng µg-1) and cg (ng m-3) are taken to be the equilibrium gas- and particle-phase concentrations, respectively. Within a given compound class and for atmospheric particles of a given type, compound-dependent Kp values can usually be correlated using the corresponding values of the compound-dependent subcooled liquid vapor pressure p°L (Torr) (1-3). In the general case, we have

log Kp ) mr log p°L + br adsorptive and/or absorptive partitioning (2) where values of the slope (mr) are frequently close to -1. When partitioning occurs solely by absorption into an organic matter phase, Pankow (2) has derived

Kp )

760RTfom 6

(3)

10 MWomζomp°L

where it is assumed that it is a primarily particle-phase organic material (om) phase that is responsible for the absorptive uptake, R is the gas constant () 8.2 × 10-5 m3 atm mol-1 K-1), T is temperature (K), fom is the weight fraction of the total suspended particulate (TSP) material that constitutes the absorptive om phase (including any water therein), MWom (g mol-1) is the number average molecular weight of the om phase (including any water therein), and ζom is the mole fraction-scale activity coefficient of the compound of interest in the om phase. As discussed by Pankow (2) and Liang and Pankow (4), eq 3 indicates that Kp values for absorptive partitioning can be expected to be proportional to fom. In cases when the dominant partitioning process is distribution into an om phase, then it is meaningful to normalize by fom (5) to obtain

Kp,om )

Kp fom

(4)

where

Kp,om )

760RT 6

(5)

10 MWomζomp°L

For a given compound, two different samples of atmospheric particulate material (PM) will therefore give similar Kp,om values at a given temperature when the product MWomζom is similar for the two samples. Liang and Pankow (4) examined that hypothesis by taking Kp values measured for PAHs sorbing to ambient urban particulate material (UPM) and to environmental tobacco smoke (ETS) PM and normalized both sets of Kp values by the estimated values of fom for the two types of PM (fom ) 0.1 and ∼1.0, respectively). Good agreement was found between the two sets of Kp/fom values. Organic compounds in the atmosphere can also adsorb to particle surfaces (1, 2). Dachs and Eisenreich (6) have suggested that particulate elemental carbon (EC) in the atmosphere may provide significant surface area for the adsorption of PAHs. Since there is currently a lack of knowledge regarding the typical values of the specific surface area, aEC (m2 g-1), of EC in ambient samples, normalization of Kp values by the quantity (m2 of EC surface area)/(µg of PM) () fECaEC/106) is not currently possible. Nevertheless, in cases when adsorption to EC completely dominates 10.1021/es011048v CCC: $22.00

 2002 American Chemical Society Published on Web 11/06/2002

the partitioning, then it may be meaningful to normalize simply by fEC according to

Kp,EC )

Kp fEC

(6)

While Liang and Pankow (4) assumed a single typical fom value for the urban Kp data set they considered, ambient atmospheric PM will surely exhibit a range of fom and fEC values, especially when considering samples obtained at different types of locations (e.g., urban, rural, etc.) and under different meteorological conditions. The purpose of this work was to determine whether variations in measured Kp values for important semivolatile organic compounds (SOCs) could be explained in terms of variations in the chemical composition of the corresponding PM samples. The SOCs considered included a range of polycyclic aromatic hydrocarbons (PAHs) and polychlorinated dibenzodioxins and dibenzofurans (PCDD/Fs). The measurements were conducted using “controlled field experiments” (CFEs) under conditions of constant temperature and relative humidity (RH). The CFEs equilibrated previously collected ambient PM with gas-phase SOCs at constant concentrations that were 1000 times or more greater than the corresponding ambient gas-phase levels of those SOCs. Such CFEs will not therefore be affected by, nor allow a determination of, the effects on ambient cp/cg ratios of small amounts of non- or slowly exchanging SOCs (7). CFEs avoid the wide ranges in temperature and RH that can be experienced with conventional, multi-day, high-volume air sampling. They therefore allow measurements of the G/P partitioning to the collected PM that can be achieved on the CFE time scale utilized while eliminating the effects of filter “blow-off” and filter “blow-on” as may be driven by changes in temperature, RH, and SOC concentrations during sampling.

Experimental Section CFE Sampling. Three sites were used: (i) University of Colorado, Denver, CO (urban); (ii) OGI/OHSU, Portland, OR (suburban); and (iii) Hills Observatory, Hills, IA (rural/farm). In the first stage of a CFE, ambient particles were collected on a precleaned Teflon membrane filter (TMF) (Pall/Gelman, Ann Arbor, MI). Once an adequate amount (∼5-10 mg) of PM had been collected to allow Kp values to be measured by the CFE method, particle collection was ended. In the second stage, the particles collected on the TMF were equilibrated with SOCs using the apparatus shown in Figure 1 of ref 8. TMF precleaning was accomplished by rinsing each 20.3 cm × 25.4 cm TMF three times with 50 mL of methylene chloride. Micrometrics, Inc. (Norcross, GA) measured the specific surface area of the TMFs to be 0.21 m2 g-1. Prior to sampling, TMFs were equilibrated overnight at 65% RH and then weighed. For each CFE, particles were first collected using a front/backup pair of precleaned TMFs (or GFFs for the single, Table 1 experiment), a high-volume (HIVOL) air sampler, and a sampling period of 11-430 h. The corresponding gas-phase compounds were not collected. After TMF particle loading was complete, a 100 mm diameter punch of the front/backup TMF pair was loaded into a filter holder that was then placed in a constant-temperature environmental chamber unit (ECU) (Figure 1 of ref 8). A glass fiber filter (GFF) in a filter holder was located in front of the TMF filter holder. Ambient, outdoor air was then drawn into the system through a length of tubing housed in an air conditioning unit (ACU). Particles in the air were removed by the GFF. The ambient gas-phase organic compounds passing through the GFF were allowed to continue on through the TMF filter holder, thereby minimizing the “stripping” of such compounds from either the particles or the TMFs themselves. The ACU and ECU were set at the same temperature.

The tubing passing through the ACU was stainless steel (45 cm long with a 0.95 cm i.d.) and was located in a water bath. The temperature sensor for the water bath was located in the body of the TMF filter holder. The maximum expected outdoor temperature was the set point for the ACU water bath, the ECU, and therefore the CFE. The air leaving the ACU was characterized by that temperature and by a nearlyconstant RH. (During stable meteorological conditions, air parcels do not tend to mix vertically, and the water content expressed in ppmV will generally stay approximately constant. Under such conditions, holding the air temperature at a constant value will maintain the RH at a near-constant value.) RH was measured using an Omega CTH89 temperature/ humidity recorder (Omega Engineering, Stamford, CT). SOCs were added to the particle-free air flow upstream of the TMF filter holder by routing a small portion (5 L min-1) of the incoming air through SOC “generator cartridges”. Specific details of the design and operation of the generator cartridges are provided elsewhere (9). The particle-free flow from the generator cartridges was mixed back into the main flow at a point upstream of the TMF filter holder. After that filter holder, the air was directed through a 24-mL front PUF plug and then a 178-mL backup PUF plug. The PUF plugs allowed determination of the cg values to which the collected particles and TMFs had been exposed. PUF plugs were changed at intervals that ranged from 10 min to 20 h; the sample volumes for those intervals ranged from 0.6 to 70 m3. The total air flow rate through the TMFs was constant within every experiment but ranged from 11 to 72 L min-1 among the different experiments and depended on the particle loading on the front filter and the air temperature. The flow rate was chosen so that the pressure drop across the two TMFs was less than 0.1 atm (40 in. of water). The duration of each experiment depended on the air temperature and ranged from 1.5 to 21 days, using a total air volume (Vt) between 55 and 920 m3. Colder temperatures required longer experimental times because the G/P and G/TMF partition coefficients were correspondingly larger. The achievement of G/TMF partitioning equilibrium was important because of the need to correct for the adsorption of gaseous SOCs to the front TMF (see Mader and Pankow (8)). At the point that an experiment was ended, it was considered likely that G/P and G/TMF equilibrium had been achieved if, for all SOCs of interest: (i) the gas-phase concentration exiting the filter (as measured with the PUF plugs) had approached an asymptotic value (designated as cg,eq); and (ii) Vt was at least twice the volume required to deliver the total mass of the compound that was found on the two TMFs, that is

Vt g 2(mf + mb)/cg,eq

(7)

where mf and mb are the amounts of the compound found on the front and backup filters, respectively. Equation 7 assumes that the gas concentration of each compound entering the filter equaled cg,eq for that compound for the entire duration of the experiment. HIVOL Sampling. At the same time that PM was being collected on a TMF for the CFE, ambient sampling was conducted using a conventional HIVOL air sampler. Those samples were used to obtain some conventional G/P partitioning data and concurrent PM-phase organic carbon (OC) and elemental carbon (EC) levels. A standard HIVOL air sampler (Graseby Manufacturing, Village of Cleves, OH) was used with front and backup 20.3 × 25.4 cm quartz fiber filters (QFFs) (Pall/Gelman, Ann Arbor MI). Corrections for gas adsorption to each front QFF were made by using analyses of the corresponding backup QFF. Prior to sampling, each QFF was baked at 450 °C for 4 h, equilibrated overnight at 65% RH, weighed, and then loaded into a filter holder. Most VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5219

sampling was conducted at times of the day during which the temperature fluctuations were relatively small. The HIVOL sample flow rate was 1 m3 min-1, yielding sample volumes of 360-720 m3. Carbon Measurements. The amounts of inorganic carbon, OC, and EC on the punches from the particle-loaded, backup, and blank QFFs were determined by Sunset Labs (Forest Grove, OR) using a thermal optical carbon analyzer (10). Particle-phase OC values were corrected for adsorption of gaseous OC using backup filters. On the basis of the approach of Mader and Pankow (11), data were only accepted for those sampling events for which it was likely that the backup filter had reached equilibrium with the incoming gaseous OC (i.e., the sample volume was large and/or the RH was high). Filter Extractions and Analyses. Immediately after an experiment, each TMF was spiked with aliquots of four surrogate standard solutions: 4 µL of 250 ng µL-1 perdeuterated fluorene in hexane; 4 µL of 250 ng µL-1 perdeuterated pyrene in hexane; 16 µL of 50 ng µL-1 fully 13C-labeled 1234TCDF in toluene; and 16 µL of 50 ng µL-1 fully 13C-labeled 2,7+2,8-monoCDD in toluene. Each particle-loaded front TMF was then Soxhlet extracted for 24 h with 100 mL of methylene chloride. Each backup TMF was extracted four times with 25 mL of methylene chloride using 10 min of sonication each time. Each extract was concentrated to 2 mL using a rotoevaporator (Buchi Instruments, Switzerland). Each extract from a particle-loaded front TMF or QFF was dried and cleaned up on a mini-column made from a disposable pipet filled with 0.2 g of silica gel and an overlying layer of 1 g of sodium sulfate. No cleanup step was necessary for the backup TMFs and QFFs. Each extract was transferred to a precleaned 4-mL vial and stored at -20 °C. Immediately before analysis by GC/MS, each extract was gently blown down to 200 µL using a stream of ultraclean N2 and then spiked with 2000 ng of perdeuterated phenanthrene. All deuterated and 13C-labeled compounds were from Cambridge Isotope Labs (Andover, MA). The extracts were analyzed on an HP 5890/5971 GC/MS using a 30 m × 0.25 mm, 0.25 µm film thickness DB-5 fusedsilica capillary column (J&W Scientific Folsom, CA). Each extract was injected “splitless” with the injector at 280 °C. The GC temperature program was 100-200 °C at 4 °C min-1; 200-250 °C at 5 °C min-1; isothermal at 250 °C for 5 min. The MS was operated in electron impact mode. The PAH standard solution contained fluorene, fluorene-d10, phenanthrene, phenanthrene-d10, anthracene, fluoranthene, pyrene, pyrened10, benz[a]anthracene, and chrysene. The PCDD/F standard solution contained 2-monoCDF, 28-diCDF, 246-triCDF, 238triCDF, 1378-TCDF, 2378-TCDF, 12378-PeCDF, 23478PeCDF, 1-monoCDD, 2-monoCDD, 23-diCDD, 28-diCDD, 124-triCDD, 1234-TCDD, 2378-TCDD, 12478-PeCDD, 12378PeCDD, and the two fully 13C-labeled internal standard compounds (all from Cambridge Isotopes Labs). Response factors for the PAHs and PCDD/Fs were determined as a function of mass injected. The internal standard compound for the PAHs was phenanthrene-d10. The internal standard compounds for the PCDD/Fs were the 13C-labeled PCDD/ Fs. PUF Extractions and Analyses. Each PUF plug was extracted using a flow-through extraction method (12). Each front and backup PUF plug was loaded into a glass syringe with a volume of 10 and 100 mL, respectively. Each PUF plug was then spiked with an aliquot of each surrogate standard solution and extracted with methylene chloride. The front and backup PUF plugs were extracted with 40 and 100 mL of methylene chloride, respectively. The extracts were reduced to 4 mL using rotary evaporation and stored at -20 °C until analyzed by GC/MS. No cleanup was necessary. N2-blow down, addition of the internal standards, and analysis by GC/MS proceeded as with the filter extracts. 5220

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002

TABLE 1. Organic Carbon (OC) and Elemental Carbon (EC) before and after a CFE Conducted with QFFs at the Beaverton Site on June 29-July 1, 1998a front filter at start of CFE front filter at end of CFE backup filter at end of CFE

OC (µg cm-2)

EC (µg cm-2)

46.8 ((2.81) 45.1 ((2.71) 4.2 ((0.63)

12.8 ((0.77) 11.8 ((0.71) 0b

a Sampling conditions: temperature ) 26 °C, RH ) 40%, and sampling time ) 44 h. b Not significantly greater than blank value of 0.1 µg cm-2.

QA/QC. Blank TMFs, QFFs, and PUF plugs were extracted and analyzed regularly. Except for chrysene, the blank level for each compound was always less than 5% of the total mass measured on the front filters and PUF plugs. For chrysene, when the blank amount corresponded to more than 10% of the sample amount, a Kp value was not computed for that sample run. For the PAHs, absolute recoveries from the PUF plugs and TMFs averaged 105%, and 107%, respectively. For the PCDD/Fs, absolute recoveries from the PUF plugs and TMFs averaged 63% and 92%, respectively. For the PUF plugs, at the warmest CFE temperature used (26 °C), breakthrough from the front to backup plug averaged 60, 5, and 7% for fluorene, phenanthrene, and anthracene, respectively, and 15, 8, and 9% for 1-monoCDF, 1-monoCDD, and 2-monoCDD, respectively. Breakthrough was negligible for the other compounds at 26 °C. Since the volume of the backup PUF plug (178 mL) was over seven times that of the front PUF plug (24 mL), when combined, it seems certain that together the two plugs provided quantitative recoveries for all compounds at all experimental temperatures. Degree of Constancy of Filter OC Levels during CFEs. Cotham and Bidleman (13) observed that passing clean air through a particle-loaded GFF can strip OC from a filter. As noted above, to minimize this effect on the particle-loaded TMFs used here, clean air was not used in the CFEs, but rather particle-free air that still contained the gaseous, ambient organic compounds. To test whether OC was conserved during CFEs, PM was collected at the Beaverton site starting on June 29, 1998, using a 20.3 × 25.4 cm QFF. Once the amount of PM that had been collected was large enough (∼5-10 mg) to allow Kp values to be measured by the CFE method, two 1.45-cm2 punches were immediately taken for OC and EC determinations. A 78.5-cm2 punch from the same particle-loaded QFF was then taken and used as the front filter in a CFE with a 78.5-cm2 punch of a clean QFF used as a backup filter. The OC and EC contents of both 78.5-cm2 QFF punches were measured upon completion of the CFE.

Results and Discussion Degree of Constancy of Front Filter OC Levels during a CFE. For the CFE conducted in Beaverton on June 29, 1998, using QFFs, the difference between the post- and pre-CFE front filter values for OC was only -4%. The corresponding difference between the EC values was -8% (Table 1). These small differences are probably not statistically significant. Nevertheless, if they were real, given that the magnitude of the percentage difference for EC was larger than that for the OC, the essential nonvolatility of EC would suggest that there may have been a slightly uneven loading of PM on the QFF and not an actual loss of OC during the CFE. Ambient OC and EC Concentrations. The ambient OC and EC data are given in Table 2. For ambient OC at the Beaverton site, there was a statistically significant (95% confidence level) difference between the average OC concentration for sunny atmospheric conditions (5.6 ( 2.2 µg m-3) vs that for rain events (2.1 ( 1.0 µg m-3). The Beaverton OC value for sunny conditions was lower than the OC

TABLE 2. Sampling Conditions and Particulate Material Characteristics for HIVOL and CFE Sampling location

date

conditions

TSP (µg m-3)

tamb a (°C)

∆tamb b (°C)

RHamb (%)

tCFE (°C)

RHCFE

fOC

fEC

OC/EC

Beaverton Beaverton Beaverton Beaverton Beaverton Beaverton Beaverton Beaverton Beaverton Denver Iowad Iowa

5/1/98 5/19/98 6/29/98 6/29/98 7/21/98 10/01/98 12/1/98 1/20/99 3/5/99 5/23/99 8/23/99 8/26/99

sunny rain sunny sunny sunny sunny rain rain rain sunny sunny sunny

96 12 15 27 43 32 15 10 22 110 20 300

19 13 13 13 17 12 9 7 4 25 16 26

-14 0 nac 0 -6 -1 -2 na 0 6.2 4.4 -5.3

58 100 na 86 81 95 89 96 89 25 96 86

27 21 26 26 26 26 12 13 12 25 na 25

34 45 40 40 40 35 46 41 50 25 na 56

0.099 0.187 0.200 0.200 0.138 0.147 0.153 0.089 0.142 0.162 0.157 0.018

0.016 0.164 0.055 0.055 0.019 0.046 0.262 0.097 0.052 0.027 0.016 0.000

6.19 1.14 3.64 3.64 7.26 3.20 0.58 0.92 2.73 6.00 9.81 na

a Mean ambient temperature. b Temperature range during sampling (a negative value indicates decreasing temperature during sampling). c na, not available. d CFE not conducted on this day.

value for downtown Denver (12.5 ( 5.5 µg m-3) and slightly greater than the average OC value measured in rural Iowa (4.3 ( 1.1 µg m-3). For ambient EC levels, a two-tailed t-test indicates a 78% probability that the average concentration measured in Beaverton during sunny conditions (1.3 ( 0.3 µg m-3) was the same as that measured for rain events (2.0 ( 1.3 µg m-3). Both of these EC values are lower than the average EC value measured for sunny conditions in Denver (2.8 ( 0.2 µg m-3) and greater than the average EC value measured for sunny conditions in Iowa (0.16 ( 0.2 µg m-3). These data are consistent with the urban, suburban, and rural natures of the Denver, Beaverton, and Iowa sampling locations. Kp Values by CFE. The gas-phase concentration of each SOC in the air exiting the TMF filter holder approached a constant asymptotic value of cg,eq as each CFE proceeded (e.g., see Figure 2 of ref 8). For the PAHs, each cg,eq was computed as the average of the measured cg values once cg varied by no more than (10%. For the PCDD/Fs, the criterion used was variation by no more than (20%. Values of Kp are reported here only for those cases when the eq 7 criterion was satisfied. Figure 1 gives plots of log Kp versus log p°L as measured by the CFE method for the three locations. The p°L values for the PAHs were from Yamasaki et al. (14); the p°L values for the PCDD/Fs were from Mader and Pankow (9). The slope and intercept values for the regression lines are given in Table 3. Also plotted in Figure 1 are previously measured log Kp versus log p°L data for partitioning of PAHs to diesel PM (15), to ETS PM (4), and to chamber-generated secondary organic aerosol (SOA) (5). While those additionally plotted data were obtained using conventional air sampling methods, they were acquired using artifact minimization techniques (e.g., sampling when the variations in temperature and RH were small and using backup filters to correct for gas adsorption). We also note that the TSP values measured for those three types of PM samples were on the order of 400 µg m-3 and that Mader and Pankow (11) discuss that gas adsorption artifacts can quickly become small for cp under such high TSP conditions. For PAHs with similar log p°L values, there is a 2 order of magnitude range in the log Kp values in Figure 1a. For just the CFE data, the Denver and Beaverton locations yielded log Kp values for the PAHs that were statistically the same over the range of log p°L values considered (95% confidence level); the Iowa log Kp values were significantly lower. The site-to-site comparisons for the PCDD/Fs data yield analogous conclusions (Figure 1b). Last, at each of the three individual CFE sampling locations, for a given value of log p°L, the PAHs and PCDD/Fs exhibited values of log Kp that were essentially the same (95% confidence level); Figure 2 presents the Beaverton data.

FIGURE 1. Measured log Kp vs log p°L by CFE at three locations for (a) PAHs and (b) PCDD/Fs. Normalizing Kp by fom. Liang and Pankow (4) and Liang et al. (5) have suggested that the G/P partitioning of PAHs and n-alkanes to UPM may be predominately absorptive in nature and describable by eq 4. Analogously, it has been observed that soil/water and sediment/water partitioning of hydrophobic organic compounds can take place into the organic material that is present in soils and sediments (16, 17). VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5221

FIGURE 2. Measured log Kp vs log p°L by CFE for PAHs and PCDD/Fs at Beaverton, OR. Since organic matter consists partly of organic carbon, for a given atmospheric PM sample, fom is expected to be related to the mass fraction of the TSP that is organic carbon (fOC). The fOC values for the sampling events of this study were obtained from the HIVOL OC measurements (Table 2). We emphasize that (i) analogously with fom, the fOC and fEC values discussed here refer to mass fractions in the TSP, with the water in the TSP included the weight of the TSP; (ii) for the measurements made in the current study, the fom, fOC, and fEC values pertain to aerosol at equilibrium with RH ≈65%; (iii) in contrast, the soils/sediments literature defines foc as the dry-weight fraction of organic carbon in the soil or sediment. The fom values for the PM samples obtained in this study were estimated as follows. For a given sample of PM, there will be a factor φw according to which

fom ) φw fOC

(8)

wherein the subscript w for “wet” emphasizes that the om phase contains some water. The factor φw will depend on RH as well as the composition of the PM. In cases when the partitioning is predominated by distribution into an om phase, then from eq 4 we have

Kp,om )

Kp φw fOC

(9)

A relation between fOC and the fraction fom,d of organic material in PM on a dry-weight basis may be written as

fom,d ) φd fOC

(10)

For primary PM, Countess et al. (18) have estimated that φd ≈ 1.2. For secondary PM, assuming an organic material density of 1.0 g cm-3, Izumi and Fukuyama (19) estimated that φd ≈ 2. For the prolonged rain events sampled in Beaverton, we will assume that only primary PM was collected since during such a rain event the residence time of particles in the atmosphere is likely sufficiently short that the formation of secondary organic aerosol is minimal. Although some water was undoubtedly present in the PM collected during those events, in the absence of any information on differences between φw and φd, we will assume that φw ≈ 1.2. Denver PM was collected near sources of primary particles (e.g., roadways) but also under sunny atmospheric conditions so that some secondary organic PM was likely also present. Liang 5222

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002

FIGURE 3. Estimated log Kp/fom vs log p°L. (a) PAHs by CFE at three locations and by conventional HIVOL sampling for SOA, diesel, and ETS PM. (b) PCDD/Fs by CFE at three locations. et al. (5) sampled UPM under similar conditions in Los Angeles and estimated that φw ≈ 1.6 for that UPM; that estimation was adopted here for the Denver samples. For the nonrain Beaverton events and the Iowa events, we will assume that the PM was ∼30% primary and ∼70% secondary, suggesting that φw ≈ 1.8. The SOA data considered here was obtained in a chamber (5) using initially particle-free air, and we estimate for that PM that fom ≈ 1. For the ETS samples, Liang and Pankow (4) utilized the thermooptical method of Birch and Cary (10) to estimate that 97% of the carbon was OC and 3% of the carbon was EC and that together OC+EC accounted for 71% of the PM mass; the remaining mass likely consisted mostly of water, oxygen, nitrogen, and hydrogen; we have adopted their assumption that fom ≈ 1 for that PM. For the diesel PM, on the basis of literature data reported in Birch and Cary (10) and measurements of Schauer et al. (20), we estimate that fOC ≈ 0.3. Assuming φw ≈ 1.2 for primary particles, we estimate for the diesel PM that fom ≈ 0.4. The calculated log Kp/fom values for the CFEs obtained by application of the above estimates of fom are plotted versus log p°L in Figure 3. The slope and intercept values for the regression lines are given in Table 3. When Figure 3 is compared with Figure 1, it is seen that normalizing the Kp values by the estimates of fom reduces the degree of variation in the partitioning data for both the PAHs and the PCDD/Fs.

TABLE 3. Regression Parameters for Log Kp and Normalized Log Kp Values vs Log p°L for CFE Data for PAHs and PCDD/Fsa log Kp

log Kp/fom

log Kp/fOC

log Kp/(fOC + fEC)

log Kp/fEC

mp

bp

r2

mp,om

bp,om

r2

mp,OC

bp,OC

Beaverton Denver Iowa all sites pooled

-1.12 -1.00 -1.07 -1.09

-8.62 -8.13 -9.22 -8.67

0.970 0.978 0.992 0.928

-1.13 -1.01 -1.07 -1.11

-8.00 -7.61 -7.73 -7.91

0.957 0.985 0.992 0.963

PAHs -1.12 -1.00 -1.07 -1.10

-7.75 -7.40 -7.68 -7.66

0.966 -1.07 -7.11 0.934 0.985 -1.03 -6.86 0.981 0.992 na naa naa 0.970 -1.06 -7.05 0.940

-1.10 -1.01 -1.12 -1.09

-7.86 -7.51 -7.95 -7.81

0.974 0.987 0.992 0.976

Beaverton Denver Iowa all sites pooled

-1.13 -0.98 -1.15 -1.05

-8.58 -7.98 -9.35 -8.42

0.883 0.891 0.953 0.831

-1.18 -1.00 -1.15 -1.14

-8.06 -7.50 -7.86 -7.95

0.883 0.912 0.953 0.893

PCDD/Fs -1.14 -1.00 -1.11 -1.13

-7.76 -7.30 -7.85 -7.69

0.890 -0.99 -7.04 0.944 0.911 -1.05 -6.93 0.955 0.952 naa naa naa 0.894 -0.98 -6.85 0.913

-1.11 -1.01 -1.21 -1.10

-7.85 -7.45 -8.14 -7.78

0.928 0.925 0.965 0.926

PAHs+PCDD/Fs all sites pooled -1.08 -8.59 0.894 -1.13 -7.95 0.935 -1.12 -7.70 0.940 -1.02 -6.95 0.927

-1.10

-7.82 0.958

a

r2

mp,EC

bp,EC

r2

r2

mp,OC+EC bp,OC+EC

Fit values with the largest r in the row are shown in boldface. The p values for all regressions are