Temperature-Dependent Uptake Rates of Nonpolar Organic

supplied by Brentwood Plastics Inc. (Brentwood, MO). The tubings were twice extracted overnight by soaking in pentane. SPMDs were prepared by enclosin...
0 downloads 0 Views 93KB Size
Download PDF
Environ. Sci. Technol. 2003, 37, 361-366

Temperature-Dependent Uptake Rates of Nonpolar Organic Compounds by Semipermeable Membrane Devices and Low-Density Polyethylene Membranes KEES BOOIJ,* HANNE E. HOFMANS, COEN V. FISCHER, AND EVALINE M. VAN WEERLEE Netherlands Institute for Sea Research, P.O. Box 59, 1790 AB Texel, The Netherlands

The effect of temperature on sampling rates and sampler-water partition coefficients of semipermeable membrane devices (SPMDs) and low-density polyethylene (LDPE) strips was studied in an experimental setup under controlled flow conditions. Aqueous concentrations of chlorobenzenes, polychlorinated biphenyls (PCBs), and polyaromatic hydrocarbons (PAHs) were maintained by continuous circulation of the water over a generator column. Sampling rates for standard design SPMDs (460 cm2) were in the range of 20-200 L d-1. No significant differences were observed between sampling rates of SPMDs and LDPE strips, but the latter samplers reached equilibrium faster because of their smaller sorption capacity. Sampling rates at 30 °C were higher than at 2 °C by a factor of about 3. Sampling rate modeling indicated boundary layer-controlled uptake for compounds with log octanolwater partition coefficients smaller than 4.4 and aqueous boundary-layer controlled uptake for more hydrophobic compounds. SPMD-water partition coefficients did not significantly change with temperature, but LDPE-water partition coefficients were larger at 2 °C than at 30 °C by a factor of 2. For field application of SPMDs, the results imply that temperature is not a key factor that controls uptake rates unless large geographical and temporal scales are involved. The results confirm that water flow velocity has a profound effect on sampling rates.

Introduction Passive sampling of dissolved organic contaminants by semipermeable membrane devices (SPMDs) is widely used for screening and source identification of a variety of nonpolar and moderately polar organic contaminants (1-7). Despite the wide application of SPMDs, calibration data that relate absorbed amounts to aqueous concentrations are rare. As a result, SPMD data collected in the field are primarily interpreted in terms of absorbed amounts, and only occasionally are the absorbed amounts translated into aqueous concentrations (6, 7). At constant temperature and flow velocity, the amount of a particular compound absorbed by an SPMD is linearly proportional to the compound’s aqueous concentration. This means that the current practice of comparing absorbed amounts in SPMDs only works well for * Corresponding author phone: (+31) 222 369 463; fax: (+31) 222 319 674; e-mail: [email protected]. 10.1021/es025739i CCC: $25.00 Published on Web 12/05/2002

 2003 American Chemical Society

deployments on small geographical and temporal scales, where differences in water flow velocity and temperature are small. In addition, it is often difficult to properly interpret differences in absorbed amounts between different compounds because little is known about the compound properties that control sampling rates. Therefore, there is a need for both additional SPMD uptake calibration data and a critical assessment of the physicochemical properties that control the uptake rates. The rate of change of the analyte concentration (Cs) in SPMDs is governed by

(

dCs koA Cs ) Cw dt Vs Ksw

)

(1)

where t is time, ko is the overall mass transfer coefficient, A is the surface area, Vs is the SPMD volume, Cw is the concentration in the water phase, and Ksw is the SPMDwater partition coefficient (3). When the transport resistance in the triolein phase can be neglected, the resistance to overall mass transfer (1/ko) is given by

1 1 1 ) + ko kw kmKmw

(2)

where kw and km are the mass transfer coefficients for the aqueous boundary layer and the low-density polyethylene (LDPE) membrane, respectively, and Kmw is the membranewater partition coefficient (3, 8, 9). Uptake Models for Constant Cw. When aqueous concentrations are constant and the initial concentration in the sampler equals zero, eq 1 can be readily integrated (3, 8-10) to give

Cs ) CwKsw(1 - exp[-ket])

(3)

where the exchange rate coefficient (ke) is defined as

ke )

Ako KswVs

(4)

For short-term exposures (ket f 0), eq 3 reduces to

Ns ) CsVs ) CwkeKswVst ) CwkoAt

(5)

) CwRst where Ns is the amount of chemical sampled and eq 5 essentially defines the water sampling rate (Rs) (3, 8-10). For long-term exposures (ket f ∞):

Cs ) CwKsw

(6)

which shows that the equilibrium is attained when the exposure time is sufficiently large (3, 9). Uptake Models for Nonconstant Cw. In laboratory experiments it is often difficult to keep aqueous concentrations sufficiently constant. A relatively simple solution to eq 1 can be obtained when Cw varies linearly with time, according to

Cw ) Cw0 + C′t

(7)

where Cw0 is the aqueous concentration at t ) 0, and C′ is the concentration rate of change. Combining eqs 1 and 7 and using the method of complementary and particular VOL. 37, NO. 2, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

361

25 °C, Hofmans (22) obtained a correlation between log Dm and log M:

solutions (11), the solution of eq 1 can be found as

(

Cs ) Cw0 -

)

C′ K (1 - exp(-ket)) + C′Kswt ke sw

(8)

It is instructive to consider a number of limiting cases that indicate that eq 8 is more than just another complex formula. First, eq 8 reduces to eq 3 when C′ ) 0 (i.e., constant Cw). Second, for long exposure times (ket f ∞), eq 8 can be reduced to

(

(

Cs ) Ksw Cw0 + C′ 1 -

))

1 t ≈ Ksw(Cw0 + C′t) ket

(9)

which shows that the SPMD will eventually follow the water phase and reach its (time-dependent) equilibrium concentration (derivation: see Box A, Supporting Information). Third, for short exposure times (ket f 0), eq 8 reduces to

(

Cs ) Ksw Cw0 +

C′t kt 2 e

)

(10)

(derivation: see Box B, Supporting Information). The term between the parentheses can be identified as the timeweighted average of aqueous concentrations after an exposure time t and is an expression of the time-integrated sampling that is characteristic of SPMDs that are operated far from equilibrium. Boundary Layer-Controlled Uptake. Extrapolation of Rs values for a particular analyte to other compounds should be done with caution. For compounds with a log octanolwater partition coefficient (log Kow) > 4.5, the SPMD-water exchange kinetics are controlled by the aqueous boundary layer, i.e., ko = kw (10, 12). Generally, kw is a function of the flow intensity near the sampler-water interface, the kinematic viscosity of the water, the sampler geometry, and the compound’s aqueous diffusion coefficient Dw (13-16). Typically, kw varies with Dw2/3 (13); therefore, a similar dependence can be expected for the sampling rates for compounds under boundary layer-controlled uptake. A number of equations relate aqueous diffusion coefficients to analyte molecular size. The Sutherland-Einstein equation (17) and the Wilke-Chang equation (14, 18) predict that aqueous diffusion coefficients change with the diffusant molar volume to the power -0.33 and -0.6, respectively. Because molar volume and molecular weight (M) are closely related quantities, Worch (14) obtained Dw ∼ M-0.53 by regression of log Dw versus log M, which indicates that SPMD sampling rates can be expected to decrease with molecular weight according to

Rs ∼ M-0.35

(11)

For a typical molecular weight range of 200-450 Da, Rs would decrease by a factor of about 1.3. Membrane-Controlled Uptake. When the membranewater partition coefficient decreases, mass transfer becomes increasingly controlled by the membrane (eq 2). For membrane-controlled uptake, ko and Rs are proportional to the product of the mass transfer coefficient for the membrane (km) and the membrane-water partition coefficient (Kmw). The value of km can be calculated from the diffusion coefficient in the membrane (Dm) and the membrane thickness (δm) according to

km )

Dm δm

(12)

Molecular size is the major factor that controls diffusion coefficients in polymers, although it is recognized that diffusant shape parameters are important as well (19-21). Using five literature sources with diffusion data in LDPE at 362

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 2, 2003

log Dm ) -7.47-2.33 log M

(13)

R ) 0.84, s ) 0.44, n ) 42 where Dm is given in m2 s-1 units. Over the molecular weight range covered by eq 13 (70 < M < 630), Dm decreases sharply from 10-12 to 10-14 m2 s-1. The other factor that determines the magnitude of sampling rates under membrane control is Kmw. Very few data exist, but when membrane-water partitioning is governed by hydrophobic interactions, it can be expected that a 2-fold increase in molecular weight would be accompanied by an increase in Kmw of several orders of magnitude. This would mean that sampling rates under membrane control increase with increasing molecular weight, although this effect will be smaller than expected on the basis of Kmw alone. In the present study, we aimed to evaluate the effect of temperature on sampling rates and sampler-water partition coefficients for SPMDs and (triolein-free) LDPE strips and to better understand the relation between sampling rates and compound properties. In addition, we aimed to facilitate the translation of amounts absorbed by field-exposed SPMDs into aqueous concentrations.

Materials and Methods Solvents. All organic solvents were of picograde quality and were obtained from PromoChem (Wesel, Germany). Ultrapure water with total organic carbon concentrations 99% of the analyte amounts. A total of 12 unexposed SPMDs and 12 unexposed LDPE strips were processed as blanks. Quantitation limits (mean + 10 × standard deviations of the blank values) were determined as 1 ng for chlorobenzenes and PCBs and 2 ng for PAHs, except for phenanthrene (17 ng). Absorbed amounts of CB209 could not be quantified in any of the samplers. Benzo[g,h,i]perylene and indeno[1,2,3-c,d]pyrene could only be quantified in the samplers that were exposed at 30 °C. Water samples (4.8 L) from the exposure system were siphoned off through stainless steel tubing into a 5-L volumetric flask. The water was extracted with about 100 mL of pentane for 30 min using a magnetic stirrer and an uncoated stirrer bar. Solvent blanks were processed together with each water sample. Quantitation limits of dissolved analytes, calculated for the individual compounds, amounted to about 0.02 ng L-1 for CB194 and CB209, 0.5 ng L-1 for chlorobenzenes and the other PCBs, 3 ng L-1 for phenanthrene, and 0.2 ng L-1 for the other PAHs. Aqueous concentrations of CB209, benzo[g,h,i]perylene, and indeno[1,2,3-c,d]pyrene always fell below the quantitation limits for these compounds, and the same was true for benzo[e]pyrene in the 2 °C experiment. The concentrated extracts of SPMDs, LDPE strips, and water samples were passed through a 6 mm i.d. LC column that contained 2 g of silica (0.063-0.200 mm, deactivated with 6% water, elution with 40 mL of pentane). The eluate

FIGURE 2. Time evolution of aqueous concentrations of pyrene during the exposure experiments at 2 (circles), 13 (squares), and 30 °C (triangles). was concentrated to 0.1 or 1 mL, followed by analysis for PCBs, chlorobenzenes, and PAHs. The cleanup procedure was sufficient to quantitatively retain 200 mg of 95% triolein (less than 0.01 mg of residue in the eluate), which indicates that both triolein and its impurities (oleic acid, methyl oleate) were sufficiently removed. Polyethylene waxes in the samples (∼1 mg) most likely were largely retained in the injection liner, which was replaced on a regular basis. No signs of column deterioration due to the precipitation of LDPE waxes was apparent. Samples were analyzed for PCBs and chlorobenzenes on a Carlo-Erba 5160 gas chromatograph equipped with a A200S autosampler, a CP-Sil 8 capillary column (50 m, 0.25 mm i.d., film thickness 0.25 µm, carrier gas H2) and an electron capture detector, model ECD-80 (constant-current mode, 340 °C). Injections were made in the splitless mode. The temperature program was 90 °C (4 min), increase at 15 °C min-1 to 300 °C (15 min). PAHs were analyzed on a Hewlett-Packard 5890 gas chromatograph equipped with a Hewlett-Packard 7673 autosampler and a CP-Sil 8 capillary column (50 m, 0.25 mm i.d., film thickness 0.25 µm, carrier gas He). The injection mode was splitless. The temperature program was 100 °C (0 min), increase at 15 °C min-1 to 200 °C, increase at 6 °C min-1 to 320 °C (0 min). Analytes were detected with a Fisons Auto-Spec Ultima mass spectrometer, ionization mode EI (70 eV, source temperature 250 °C), detection mode selected ion recording of the molecular ions (dwell time 80 ms, equilibration 10 ms). Data Analysis. The time evolution of aqueous concentrations was separately modeled for each compound and all experiments by linear regression (eq 7). Equations 8 and 10 were then used to model the evolution of analyte concentrations in the samplers, by nonlinear parameter estimation, assuming a log-normal distribution of errors. The simpler eq 10 (one adjustable parameter, the product keKmw or keKsw) was favored over the more complex eq 8 (two adjustable parameters, ke and Kmw or Ksw) when the complex model did not give a significantly better fit, as evidenced by the extra sum of squares method (12). Sampling rates could be calculated from the product keKsw, when either the simpler model or the more complex model applied, but a separate estimation of ke and Ksw could only be made when the application of the complex model was justified.

Results and Discussion Stability of the Aqueous Concentrations. A typical example of the time evolution of aqueous concentrations (as determined by the batch extractions) is shown for pyrene in Figure 2. The resulting pyrene absorption by SPMDs is shown in Figure 3. All aqueous concentrations and amounts absorbed VOL. 37, NO. 2, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

363

4 (Table E in Supporting Information). Comparison of a number of linear regression models that allowed for temperature-dependent intercepts and slopes showed that the log Ksw data at all temperatures could best be described by a single linear equation in log Kow:

log Ksw ) 0.715 log Kow + 1.29

(14)

R ) 0.96, s ) 0.14, n ) 26

FIGURE 3. Time evolution of pyrene concentrations in SPMDs exposed at 2 (circles), 13 (squares), and 30 °C (triangles). by the samplers are listed in Tables B-D in Supporting Information. Aqueous concentrations could adequately be described by eq 7, with the exception of hexachlorobenzene and CB28 in the 13 °C experiment, where an appreciable nonlinear time dependence was observed. Therefore, these compounds were omitted for the 13 °C experiment. The main reason for the concentration increase with time is that the uptake rate of the samplers often was of the same order as the delivery rate of the generator column. The volume rate of flow from the generator column was about 900 L d-1. The sampling rate of each SPMD and LDPE strip during the 30 °C experiment was about 60 L d-1 for each device, and during the first day of the exposure, a total of 10 devices was present in the setup. This means that the total sampling rate was about 600 L d-1. During the course of the experiments, the number of samplers in the setup was successively decreased from 10 to 2. The total sampling rate of the devices therefore dropped from 600 to 120 L d-1, thereby enabling the generator column to better keep up with the contaminant removal by the samplers, which is reflected by the increase in aqueous concentrations. It can therefore be concluded that the generator column method as such is suitable for stabilizing the aqueous concentrations, but that the total volume rate of flow from the columns should have been higher. Sampler-Water Partition Coefficients. LDPE-water partition coefficients (Kmw) and SPMD-water partition coefficients (Ksw) are shown as a function of log Kow in Figure

The present Ksw estimates for acenaphthene and phenanthrene are similar to the values reported in the literature (0.02 and 0.09 log unit difference, respectively), but the present estimates for fluoranthene and pyrene are larger by 0.44 log unit (8, 10). The observed temperature independence of Ksw values is in line with similar observations made by Huckins et al. (8). The log Kmw data could best be described as a linear function of log Kow, with a temperature-dependent intercept:

log Kmw ) 0.972 log Kow + constant

(15)

@ 2 °C: constant ) 0.13 @ 13 °C: constant ) -0.13 @ 30 °C: constant ) -0.22 R ) 0.95, s ) 0.23, n ) 32 The inclusion of a quadratic term in eqs 14 and 15 did not yield significantly better fits. Differences between Ksw and Kmw are smaller than 0.6 log unit in the range 4 < log Kow < 6. This observation suggests that the triolein-LDPE ratio in SPMDs has no large effect on the SPMD-water partition coefficient. Sampling Rates. No major differences between the sampling rates of SPMDs and LDPE strips were observed (Table F in Supporting Information). The Rs ratio for the two samplers is shown in Figure 5. The scatter around the reference line (ratio of sampling rates ) 1) amounts to a factor of 1.24, which is fairly small considering the fact that the errors in both sampling rates contribute to the scatter in their ratio. Because sampling rates are linearly proportional to the SPMD surface area, Rs values observed for the 152 cm2 SPMDs used in this study were multiplied by 460/152 in order to facilitate comparison with the more commonly used 460 cm2 SPMDs (9). Average sampling rates for the two samplers are shown as a function of log Kow in Figure 6. Rs estimates initially

FIGURE 4. LDPE-water partition coefficients (Kmw, panel a) and SPMD-water partition coefficients (Ksw, panel b) as a function of log Kow at 2 (circles), 13 (squares), and 30 °C (triangles). 364

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 2, 2003

FIGURE 5. Ratio of the sampling rates of LDPE strips and SPMDs as a function of log Kow at 2 (circles), 13 (squares), and 30 °C (triangles). increase with increasing hydrophobicity, reach a maximum at about log Kow ) 6, and then show a modest decrease up to log Kow ) 7.5. On average, sampling rates were higher at 30 °C than at 2 °C by a factor of 2.8, which is of the same order as the 1.5-fold increase in PAH sampling rates between 10 and 26 °C, reported by Huckins et al. (10). The present sampling rates of 20-200 L d-1, obtained at a water flow velocity of about 90 cm s-1, are significantly larger than reported PAH sampling rates of 2-8 L d-1 at a flow velocity of about 0.004 cm s-1 (10). Vrana and Schu ¨u ¨ rmann (23) observed an increase of the sampling rates of penta- and hexachlorobenzene from 4 to 14 L d-1, when the (laminar) flow increased from 0.06 to 0.3 cm s-1. For an additional increase in flow velocity to 1.1 cm s-1, they did not observe an increase in sampling rates, but they speculated that the transition of laminar to turbulent flow would cause the sampling rates to rise again with the 0.8-0.9 power of the bulk water flow velocity. Considering that the flow rates in the present study were higher by a factor of about 80, a further increase of sampling rates by a factor of 14 is in agreement with this prediction. The initial sharp increase of sampling rates with increasing log Kow is indicative of membrane-controlled uptake, which is rather difficult to model, since Kmw is hydrophobicityrelated, but km is related to molecular size. Within a single compound class, good correlations can be obtained between molecular weight and log Kow, but when different compound classes are involved, the quality of these correlations can be no better than fair. A reference equation for membranecontrolled sampling rates was established by calculating the value of kmKmw at 25 °C (eqs 12, 13, and 15) for the individual compounds and then regressing log kmKmw versus log Kow. The results are given by

log kmKmw ) -7.47 + 0.682 log Kow

(16)

R ) 0.99, s ) 0.10, n ) 17 where km is given in m s-1 units. After back-transformation, eq 16 gives

kmKmw ) BmK0.691 ow The coefficient Bm equals 34 nm s-1 at 25 °C, but it may attain different values at other temperatures since both Kmw and km are temperature dependent. The relative constancy of sampling rates at the high log Kow side of Figure 6 is indicative of aqueous boundary layercontrolled uptake. A complicating factor in estimating mass

FIGURE 6. Sampling rates of 460 cm2 LDPE strips and SPMDs as a function of log Kow at 2 (circles), 13 (squares), and 30 °C (triangles). Model estimates (eq 20) are shown as drawn lines.

TABLE 1. Summary of Results Obtained by Fitting Eq 20 to the Observed Sampling Rates parameter

2 °C

13 °C

30 °C

Bw (µm s-1) Bm (nm s-1)

29 ( 2 14 ( 3

48 ( 3 46 ( 6

77 ( 5 50 ( 13

transfer coefficients for the aqueous boundary layer from first principles is that the structure of the flow is generally unknown, except for some special cases such as tangential flow over semi-infinite flat surfaces and flow through straight pipes. There is general agreement in the engineering literature, however, that kw increases with the two-thirds power of the aqueous diffusion coefficient (13, 14, 16), which allows for predicting the difference in sampling rates between compounds. A model equation for log Dw was obtained by linear regression against log Kow, using Dw estimates from the Worch equation (14) evaluated for a temperature of 13 °C:

log Dw ) -8.96 - 0.0659 log Kow

(18)

R ) 0.94, s ) 0.02, n ) 17 where Dw is in m2 s-1 units. Considering that kw ∼ D2/3, the log Kow dependence of kw can be written as

kw ) BwK-0.044 ow

(19)

The coefficient Bw cannot be calculated a priori for the reasons mentioned above, but Bw is likely to be temperaturedependent mainly because the kinematic viscosity of water has a strong temperature dependence. Combining eqs 2, 5, 17, and 19, the log Kow dependence of the sampling rates can be modeled by

Rs ) koA )

A A ) (20) 1 1 1 1 + + -0.044 0.682 kw kmKmw B K BmKow w ow

Equation 20 was fitted to the data, assuming a log-normal distribution of errors and allowing for temperature-dependent coefficients Bm and Bw. The results are shown as drawn lines in Figure 6 and are summarized in Table 1. The residual errors amounted to 0.09 log unit. A second model in which the slope of the log kw versus log Kow correlation was not VOL. 37, NO. 2, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

365

fixed to -0.044, but was freely adjustable, did not give a significantly better fit. Activation energies for contaminant uptake can be estimated by plotting log Bm and log Bw versus reciprocal absolute temperature, because Bm and Bw are linearly proportional to km and kw, respectively. Values of 30 and 24 kJ mol-1 were obtained for membrane-controlled uptake and boundary layer-controlled uptake, respectively. These values indicate that every 10 °C temperature increase results in a 1.5-fold increase in sampling rate. The Bm estimate of 50 nm s-1 at 30 °C is quite close to the independent estimate of 34 nm s-1 at 25 °C (eq 17). Implications for Field Deployments. The results indicate that, even for a temperature difference of 20 °C, changes in contaminant uptake rates by SPMDs are limited to a factor of about 2. This means that differences in absorbed amounts generally cannot be attributed to temperature differences, except when very large geographic scales or summer/winter comparisons are involved. The temperature effect is even less for compounds that reach equilibrium during the deployment because we did not find a temperature effect on SPMD-water partition coefficients. By contrast, differences in water flow velocities can cause sampling rates to range from about 4 L d-1 at flows in the sub-mm s-1 range (10, 23), via 10 L d-1 in the cm s-1 range (23), to 200 L d-1 at flow velocities of about 1 m s-1. Because of the complexity of the hydrodynamics involved, particularly in the case of caged SPMDs, there is little hope that sampling rates can be expressed as a simple function of ambient flow rates. Estimating in situ sampling rates by measuring the dissipation rates of performance reference compounds (PRCs) (2, 8, 9, 12) is therefore mandatory, unless flow velocity measurements indicate that differences in the hydrodynamical conditions between the various deployment sites are minor. Sampling rates for compounds under boundary layercontrolled uptake are proportional to D2/3 w , which may be approximated by M-0.35. This relationship allows for calculating uptake rates for hydrophobic compounds from the dissipation rates of less hydrophobic performance reference compounds. For sampling rates in the order of 100 L d-1 (present study), the uptake is for 90% controlled by the aqueous boundary layer at log Kow g 5.7. Under these circumstances, the dissipation of CB29 and perdeuterated chrysene (log Kow ≈ 5.9) may be used to calculate Rs values for more hydrophobic compounds. The deployment period should be large enough, however, to ensure that the dissipated PRC amount is sufficiently large to be quantified. At low-flow conditions, the uptake is boundary layercontrolled for log Kow g 5 (10), indicating that in this case CB4 (log Kow ) 5) and perdeuterated pyrene and fluoranthene (log Kow ) 5.2) are suitable PRCs for estimating sampling rates for more hydrophobic compounds.

Acknowledgments This study was supported by a grant from the National Institute for Coastal and Marine Management/RIKZ under Contract RKZ 583. This is NIOZ Publication No. 3587.

Supporting Information Available Aqueous concentrations, sampler-absorbed amounts, sampler-water partition coefficients, sampling rates, and deri-

366

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 2, 2003

vations of eqs 9 and 10. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Huckins, J. N.; Tubergen, M. W.; Manuweera, G. K. Chemosphere 1990, 20, 533-552. (2) Prest, H. F.; Jarman, W. M.; Weismuller, T.; Martin, M.; Huckins, J. N. Chemosphere 1992, 25, 1811-1823. (3) Huckins, J. N.; Manuweera, G. K.; Petty, J. D.; Mackay, D.; Lebo, J. A. Environ. Sci. Technol. 1993, 27, 2489-2496. (4) Herve, S.; Prest, H. F.; Heinonen, P.; Hyo¨tyla¨inen, T.; Koistinen, J.; Paasivirta, J. Environ. Sci. Pollut. Res. Int. 1995, 2, 24-29. (5) Prest, H. F.; Huckins, J. N.; Petty, J. D.; Herve, S.; Paasivirta, J.; Heinonen, P. Mar. Pollut. Bull. 1995, 31, 306-312. (6) McCarthy, J. F.; Southworth, G. R.; Ham, K. D.; Palmer, J. F. Environ. Toxicol. Chem. 2000, 19, 352-359. (7) Booij, K.; Zegers, B. N.; Boon, J. P. Chemosphere 2002, 46, 683688. (8) Huckins, J. N.; Petty, J. D.; Lebo, J. A.; Almeida, F. V.; Booij, K.; Alvarez, D. A.; Cranor, W. L.; Clark, R. C.; Mogensen, B. B. Environ. Sci. Technol. 2002, 36, 85-91. (9) Huckins, J. N.; Petty, J. D.; Prest, H. F.; Clark, R. C.; Alvarez, D. A.; Orazio, C. E.; Lebo, J. A.; Cranor, W. L.; Johnson, B. T. A Guide for the Use of Semipermeable Membrane Devices (SPMDs) as Samplers of Waterborne Hydrophobic Organic Contaminants; API Publication 4690; American Petroleum Institute: Washington, DC, 2002; pp 1-192. (10) Huckins, J. N.; Petty, J. D.; Orazio, C. E.; Lebo, J. A.; Clark, R. C.; Gibson, V. L.; Gala, W. R.; Echols, K. R. Environ. Sci. Technol. 1999, 33, 3918-3923. (11) Handbook of Chemistry and Physics; CRC Press: Boca Raton, FL, 1983. (12) Booij, K.; Sleiderink, H. M.; Smedes, F. Environ. Toxicol. Chem. 1998, 17, 1236-1245. (13) Boudreau, B. P.; Guinasso, N. L. In The Dynamic Environment of the Ocean Floor; Fanning, K. A., Manheim, F. T., Eds.; Lexington Books: Toronto, 1982; Chapter 6. (14) Worch, E. Vom Wasser 1993, 81, 289-297. (15) Levich, V. G. Physicochemical Hydrodynamics; Prentice Hall, Inc.: Englewood Cliffs, NJ, 1962; pp 1-700. (16) Bird, R. B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenomena; John Wiley & Sons: New York, 1960; pp 1-780. (17) Gobas, F. A. P. C.; Opperhuizen, A.; Hutzinger, O. Environ. Toxicol. Chem. 1986, 5, 637-646. (18) Danckwerts, P. V. Gas-Liquid Reactions; McGraw-Hill: New York, 1970; pp 1-276. (19) Lloyd, D. R.; Meluch, T. B. In Materials Science of Synthetic Membranes; Lloyd, D. R., Ed.; ACS Symposium Series 269; American Chemical Society: Washington, DC, 1985; Chapter 3. (20) Mo¨ller, K.; Gevert, T. J. Appl. Polym. Sci. 1994, 51, 895-903. (21) Lieb, W. R.; Stein, W. D. Nature 1969, 224, 240-243. (22) Hofmans, H. E. Numerical modelling of the exchange kinetics of semipermeable membrane devices. M.Sc. Thesis, Netherlands Institute for Sea Research, Den Burg, The Netherlands, 1998; pp 1-37. (23) Vrana, B.; Schu ¨u ¨ rmann, G. Environ. Sci. Technol. 2002, 36, 290296.

Received for review April 23, 2002. Revised manuscript received October 10, 2002. Accepted November 4, 2002. ES025739I