Evidence of Bias in Air−Water Henry's Law Constants for Semivolatile

Critical Review

Evidence of Bias in Air-Water Henry’s Law Constants for Semivolatile Organic Compounds Measured by Inert Gas Stripping CHUBASHINI SHUNTHIRASINGHAM, YING DUAN LEI, AND FRANK WANIA* Department of Chemistry, and Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4

Accurate knowledge of the air-water Henry’s law constant (H) is crucial for understanding an organic compound’s environmental behavior. The inert gas stripping (IGS) method, widely used to measure H of semivolatile organic compounds (SOCs), may yield erroneously high values for compounds with a high water surface adsorption coefficient, KIA, because chemical adsorbed to the bubble surface may be transferred to the head space upon bursting at the top of the stripping column. Experiments with alkanols of variable chain length identified a KIA threshold of approximately 10-3 m, above which IGS is susceptible to this artifact. Most SOCs are predicted to have KIA values well above that threshold. IGS-determined H-values for chemicals belonging to various groups of SOCs were evaluated by comparison with H-values either calculated from reliable vapor pressure and solubility data or derived from data compilations that achieve thermodynamic consistency through optimized adjustment of measured physical-chemical property data. The investigated deviations were found to be generally consistent with what would be expected from a surface adsorption artifact. Namely, the apparent bias in IGS-determined H-values, if it occurs, (1) is positive, (2) increases with increasing size of an SOC, and (3) increases with decreasing temperature. It generally is also of a magnitude predicted using estimated KIA values. However, different studies display different KIA threshold values, beyond which the artifact becomes notable, and some studies appear to succeed in avoiding the artifact altogether. Whereas the use of aerosol traps cannot explain the absence of a surface adsorption artifact, it may be related to higher flow rates used by some investigators. For large compounds or those with more than one functional group, the predicted deviation is too large when compared to observations, suggesting that the estimated KIA values for those compounds are too high. A full quantitative understanding of the artifact requires more accurate predictions of the adsorption of SOCs to the water surface.

Introduction The Henry’s law constant (H) describes the equilibrium partitioning of neutral compounds between the gas phase * Corresponding author [email protected]. 10.1021/es062957t CCC: $37.00 Published on Web 04/28/2007

phone:

(416)

287-7225;

 2007 American Chemical Society

e-mail:

and a dilute aqueous solution. H is defined as the ratio of the partial pressure of a solute to its concentration in a dilute aqueous solution at equilibrium (1)

H ) PG/CW

(1)

where H is in Pa‚m3‚mol-1, CW is the solute concentration in mol‚m-3, and PG is the partial pressure of the solute in the gas phase in Pa. A dimensionless Henry’s law constant, or air-water partition coefficient KAW, can be derived from H by converting the partial pressure of the solute to its concentration using the ideal gas law,

KAW ) H/RT

(2)

where R is the gas constant, 8.314 Pa‚m3‚mol-1‚K-1, and T is the absolute temperature in K. For sparingly soluble organic compounds, H can be estimated as the ratio of the compound’s saturation vapor pressure P and water solubility S (1). H plays a central role in determining the importance, direction, and rate of air-water transfer processes of organic compounds (1, 2), including those of environmental interest such as the transfer processes involving cloud and rrain droplets, as well as those involving water in lakes, rivers, oceans, and the subsurface. Due to its fundamental role in controlling the environmental fate of organic compounds, it is crucial to know the H of contaminant chemicals as precisely and accurately as possible. For example, uncertain H-values are the largest source of uncertainty in calculations of the equilibrium status, and therefore in the air-water exchange, of polychlorinated biphenyls (PCBs) between air and Baltic Sea water (3). Similarly, small differences in the H of the various congeners of technical hexachlorocyclohexane (HCH) can result in widely different global fate (4). The availability of techniques for the reliable determination of H for organic chemicals of environmental concern is therefore of critical importance. The inert gas stripping (IGS) method, where chemicals are stripped from an aqueous solution into the gas phase, was first introduced by Leroi et al. (5) and applied to determine H of hydrophobic compounds by Mackay et al. (2). Briefly, in this technique inert gas bubbles through a vertical column containing an aqueous solution of the chemical of interest, whereby it has to be assured that the solute concentration found in the exit gas is in equilibrium with the aqueous concentration. The decreasing concentration C of a chemical in the exit gas or in the aqueous solution is measured as a function of time t and KAW can be calculated from the slope of the following linear relationship: VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3807

ln(C(t)/C0)) - KAWGt/V

(3)

where C0 is the concentration at time 0, G is the flow rate of the inert gas, and V is the volume of the aqueous solution. Normally the logarithm of the concentration normalized to the initial value is plotted against time, but if the decrease in V during the experiment is notable, it should be quantified and the logarithm of the normalized concentration should be plotted against t/V (6). The method only requires the quantification of concentrations in air or water, and no absolute but only relative changes in concentration need to be recorded (2). This simplicity should aid in achieving highly precise and accurate H-values. In the last 30 years, IGS has become a widely used method for experimentally determining H of semivolatile organic compounds (SOCs). For example, it has been used to measure H for PCBs (7-11), organochlorine pesticides (OCPs) (6, 7, 12-14, 15), polycyclic aromatic hydrocarbons (PAHs) (2, 7, 16-19), and polybrominated diphenyl ethers (PBDEs) (9, 20). Goss et al. (21) noted discrepancies between the H of highly chlorinated PCBs measured with IGS by Bamford et al. (8) and those determined by other techniques (22, 23). The measured H-values (8) were also not in agreement with other physical-chemical properties for these PCB congeners (21), as is evident from consistent deviations from data sets of partitioning properties for PCBs (24, 25) that adhere to thermodynamic constraints. It was hypothesized that adsorption to the surface of the gas bubbles could bias H measurements performed by IGS, if a fraction of the organic chemical adsorbed on the bubble surface is released into the headspace of the stripping column upon bursting (21). Lei et al. (26) recently demonstrated the existence of such an artifact by measuring H of normal alkanols of variable chain length at a number of temperatures using both IGS and a headspace technique that is much less susceptible to surface adsorption artifacts. For the longer-chain alkanols at lower temperatures the IGS method was found to greatly overestimate H, whereas for shorter-chain alkanols and for measurements done at higher temperatures, no such artifact due to adsorption to the water surface was observed. The artifact can be quantified with the help of an enhancement factor EF, which expresses the extent to which H from IGS is too high as a result of adsorption to the water surface (26). An EF of 1 indicates the absence of an artifact. Figure 1 displays the EF for the normal alkanols by plotting them against the logarithm of their interface-air partition coefficient, KIA, which quantifies the strength of adsorption to the water surface. KIA is defined as the ratio of the concentration on the water surface (mol‚m-2 surface) and the concentration in the gas phase (mol‚m-3 air) at equilibrium, and thus has units of m (27). Figure 1 shows that the EF is increasing rapidly if the surface adsorption coefficient of an alkanol at the experimental temperature exceeds a threshold log(KIA/m) value in the range of -3.2 and -3. Whereas EF equals 1 for less sorptive alkanols, the surface adsorption artifact can quickly result in errors of 1 order of magnitude or higher for alkanols with a log(KIA/m) of -2. Lei et al. (26) concluded that H measurements by IGS for organic chemicals with a KIA greater than 0.001 m could be susceptible to a surface adsorption artifact. No experimental KIA values for the adsorption of SOCs to the air-water interface currently exist, but we can estimate at least approximate values of KIA for PAHs, PCBs, OCPs, and PBDEs using a poly-parameter linear free energy relationship (27, for details see method section below). Figure 2 displays KIA-values for selected SOCs as a function of temperature in the environmentally relevant range of 5-35 °C (Additional SOCs with more uncertain KIA are displayed in Figure S1 in the Supporting Information). The estimated log(KIA/m) for PAHs with more than three fused rings, larger OCPs (e.g., 3808

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007

FIGURE 1. Enhancement factors indicating the extent to which H for n-alkanols measured by IGS at different temperatures (] 25 °C, 0 31 °C, 4 51 °C, O 69 °C) is too high as a result of adsorption to the water surface, plotted against the adsorption coefficient to the water surface KIA (adjusted to the experimental temperature, from ref 26). Above a log(KIA/m) threshold of -3.2 to -3 (gray vertical band), enhancement factors above 1 indicate susceptibility of the IGS to a surface sorption artifact.

FIGURE 2. Relationship between the adsorption coefficients on the air-water interface KIA, estimated for selected PAHs, OCPs, and PCBs using a poly-parameter linear free energy relationship (27), and reciprocal temperature. IGS measurements of the H of chemicals with a log(KIA/m) above the gray line may be susceptible to a surface adsorption artifact. p,p′ -DDE, p,p′-DDT, heptachlor, chlordanes), larger PCB congeners, and all PBDEs are above -3 at environmental temperaturessthe threshold at which IGS may become susceptible to a surface artifact (26). The log(KIA/m) for small OCPs, such as hexachlorobenzene (HCB), and small PAHs with less than three fused rings is less than -3.2, suggesting that these compounds sorb less strongly to the water surface and thus the IGS method should yield H values that only deviate marginally from the real value. The log(KIA/m) of some SOCs, such as fluorene, the HCHs, PCB-3, and PBC-28, passes through the threshold regions within the typical experimental temperature range. For these chemicals, we would therefore only expect a surface artifact at lower experimental temperatures. Clearly, many SOCs adsorb strongly to the water surface, and the IGS method may

overestimate the H of such chemicals. The validity of H values for such chemicals needs to be re-evaluated. This is what this study set out to do.

Theory In this paper, an H or KAW experimentally determined by IGS is called apparent Happ or KAWapp. The loss of chemical from the aqueous phase during IGS is due to the chemical partitioning from the aqueous phase into the gas phase of the bubble (controlled by KAW) and due to chemical adsorbing from the aqueous phase to the surface of the bubble (controlled by the adsorption coefficient on the water surface from the water side KIW) (26). The contribution of the bubble surface relative to the bubble’s gas phase has to be weighted by its surface-to-volume ratio and corrected for the fraction fTGB of the chemical amount on the surface that is transferred to the gas phase. The surface-to-volume ratio of a spherical bubble is (4πr2)/(4πr3/3) or 3/r, where r is the radius of the bubble. Therefore, an IGS experiment yields an apparent air-water partition coefficient of

KAWapp ) KAW + KIW‚fTGB‚3/r

(4)

Replacing KIW with KIA‚KAW, we obtain

KAWapp ) KAW(1 + fTGB‚KIA‚3/r) ) KAW‚EF

(5)

Happ ) KAW‚EF‚R‚T

(6)

EF is the enhancement factor, 1 + fTGB‚KIA‚3/r, displayed in Figure 1. Equation 5 illustrates that the larger a chemical’s KIA and the smaller the bubbles, the larger the EF and therefore the larger the positive bias in Happ. KIA increases with decreasing temperature (Figure 2), and therefore, eq 5 suggests that EF is larger at lower temperatures.

Method In this study, Happ for different classes of SOCs from IGS measurements reported in the literature, either at 25 °C or as a function of temperature, are compared with H-values that either are estimated from the ratio P/S of reliable vapor pressure P and water solubility S data from the literature (28) or are final adjusted values (FAVs) from data compilations that achieve thermodynamic consistency through optimized adjustment of measured physical-chemical property data (24, 25, 29-32). Vapor pressure and water solubility data were judged reliable if measurements by multiple investigators were in good agreement, i.e., if the deviations in P/S caused by using different sources of P and S were much smaller than the deviations between P/S and Happ. Final adjusted H-values are judged reliable because they are based on, and are thermodynamically consistent with, all available experimental data on a chemical’s partitioning between its pure liquid phase, gas phase, aqueous dissolved phase, and dissolved-in-octanol phase (P, S, KOW, KOA, KAW, solubility in octanol). Deviations between H and Happ are then evaluated whether they agree with theoretical expectations, namely whether they are of the magnitude and direction predicted by eqs 5 and 6. The Supporting Information includes a discussion of the validity of comparing Happ-values obtained by IGS with FAVs that themselves depend on these or other IGS-derived Happ. To facilitate the comparison with theoretical expectations, we estimated KIA values for SOCs using the poly-parameter linear free energy relationship by Roth et al. (27)

log(KIA/m at 15 °C) ) 0.635‚log L16 + 5.11‚B + 3.60‚A - 8.47 (7)

where L16 is the hexadecane/air partition coefficient, and A and B are measures of a solute’s hydrogen bonding acidity and basicity, respectively. L16, A, and B for benzene, selected PAHs (naphthalene, phenanthrene, anthracene, pyrene, chrysene, benzo[a]pyrene), HCHs, and the PCBs were taken from the literature(33-36), and they were estimated for PBDEs and OCPs (37). For some PAHs, L16, A, and B were estimated from linear relationships between L16, A, and B for the six PAHs mentioned above and molar volume (28). KIA was adjusted to experimental temperatures using the van’t Hoff relationship

ln (KIA(T2)/ KIA(T1)) ) (∆adsH/R) (1/T1 - 1/T2)

(8)

The enthalpy of adsorption to the water surface ∆adsH was estimated according to Roth et al. (38)

∆adsH (J/mol) ) (-5.52‚ln KIA (15 °C) -107)‚1000

(9)

There is considerable uncertainty in the estimated KIA values. First, the solute descriptors for the SOCs are generally not measured values and estimated L16, A, and B values may have significant errors. Second, eq 7 has been developed based on measured KIA values for fairly small molecules and may not be readily applicable to chemicals with more complex structures. In particular, Goss (39) stresses the importance of molecular orientation in the adsorption of molecules that have more than one functional group capable of electron donor-acceptor interactions. Such chemicals, which include many pesticides, may be sterically hindered to bring all of their e-donor and e-acceptor functional groups in contact with the complimentary functional groups of the surface, which would result in overprediction of their KIA by eq 7 (39). The value of fTGB and whether and how it depends on the properties of the bubbles and chemicals is unknown. In fact, the size of the bubbles generated in the IGS apparati used in the various studies is unknown itself. Lei et al. (26), using the original IGS apparatus by Mackay et al. (2), estimated bubble radii in the range of 0.9-1.5 mm and found a dependence of r on gas flow rate. To use eq 5 in the absence of quantitative information, various values for the ratio fTGB/r were assumed to apply.

Results and Discussion Dependence of Artifact on the Adsorption Coefficient to the Air-Water Interface KIA. Several groups have employed IGS to measure the H of individual PAHs (2, 7, 16-19) and PCB congeners (7-9, 11) at 20 or 25 °C. These data were compared with the H calculated from P and S (in the case of the PAHs) or (in the case of the PCBs) the FAVs by Schenker et al. (25), which are based on the data compilation by Li et al. (24), by plotting them as a function of log(KIA/m) (Figures 3 and 4). Happ-values measured by different groups for the PAHs and PCBs with a KIA below the threshold of 0.001 m are close to each other. In particular, there is excellent agreement between the calculated H and measured Happ for benzene and naphthalene and between the FAVs of H and the measured Happ for small PCB congeners, consistent with estimated KIA-values at 25 °C for these chemicals well below the threshold of 0.001 m (Figures 2). It thus appears that the IGS provides reliable H-values for benzene and small PAHs and PCBs, as is expected based on their low KIA-values. However, the measured Happ data for chemicals with a KIA above the threshold show large discrepancies and are not consistent with the calculated P/S ratios for the PAHs and the FAVs for the PCBs. Specifically, the Happ for PAHs such as phenanthrene and anthracene measured by Alaee et al. (16), Mackay et al. (2), Shiu and Mackay (17), and Bamford VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3809

FIGURE 3. Comparison between Happ of benzene and PAHs measured by various groups using IGS (* refs 2 and 17; 0 ref 7; ] ref 16; O ref 18; 4 ref 19), and H calculated from vapor pressure and water solubility (thick line). Data at 25 °C, except those in ref 18 which are 20 °C. The IGS yields Happ-values for larger PAHs that are higher than the P/S ratios, which is consistent with predictions of Happ using different fTGB/r ratios (thin lines).

FIGURE 4. Comparison between Happ of PCB congeners measured by various groups using IGS (4 ref 7; ] ref 8, * ref 9; ] ref 10; × ref 11), and final adjusted H-values for the PCBs (solid line, ref 25). Data at 25 °C, except those from ref 11 which are 20 °C. The IGS yields Happ-values for larger PCBs that are higher than the FAVs, which is consistent with predictions of Happ using different fTGB/r ratios (thin lines). et al. (19) are always higher than the calculated H. As the size of the PAHs increases, the Happ values reported by Bamford et al. (19) deviate considerably from the P/S ratios. Similarly, the Happ for the larger PCB congeners experimentally determined by Bamford et al. (8), Fang et al. (10), and Lau et al. (9) are higher than the final adjusted H (25) and the deviation tends to increase with increasing KIA. This is consistent with estimated KIA values for these PAHs and PCBs above the proposed threshold for a surface adsorption artifact (Figure 2) and the expectations that this artifact should increase with KIA. Figures 3 and 4 include lines showing Happ estimated with eqs 5 and 6 using the P/S ratios (PAHs) and the FAVs (PCBs) for H and assuming different fTGB/r ratios. The magnitude of the deviations between the measured Happ and H can be explained by fTGB/r ratios in the range 0.0250.25 mm-1. Such ratios correspond, for example, to the transfer to the gas phase of 2.5-25% of molecules adsorbed to gas bubbles with a 1 mm radius, or an fTGB between 10 and 100% for much larger bubbles of 4 mm radius. Difference between the Size of the Artifact in IGS Studies by Different Research Groups. The Happ for the larger PAHs 3810

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007

and PCBs measured by different groups using IGS is higher than the H judged to be correct. However, the extent of deviation between H and the Happ measured by different investigators is not the same, even though they all used the same technique. The Happ for larger PAHs determined by ten Hulscher et al. (7) and De Maagd et al. (18) deviate much less from the calculated H than those from the other IGS studies, e.g., Bamford et al. (19) (Figure 3). This is also the case for the larger PCB congeners, where the Happ values by ten Hulscher et al. (7, 11) are closer to the FAVs by Schenker et al. (25) than those reported by Bamford et al. (8), Lau et al. (9), and Fang et al. (10) (Figure 4). This can be partly explained by the fact that the Happ values reported by ten Hulscher et al. (11) apply to 20 °C, i.e., are lower because of slightly lower experimental temperatures. Figure S2 in the Supporting Information compares the Happ for PCBs measured by ten Hulscher et al. (11) with the FAVs at 20 °C by Li et al. (24), and suggests good agreement except for PCB-180. However, temperature alone cannot explain why the Happ by ten Hulscher et al. (7, 11) appear to be closer to the expected H values than the Happ reported in other IGS studies. As indicated by eq 5, Happ depends on both the KIA of a chemical and the size of the bubbles. Most IGS studies compared here used a flow rate of 200 mL‚min-1 or less, whereas ten Hulscher et al. (7, 11) and De Maagd et al. (18) reported using much higher flow rates of 500-1000 mL‚min-1. Lei et al. (26) had noted that bubble size increases with flow rate, suggesting that bubbles produced in the IGS apparatus by ten Hulscher et al. (7, 11) and De Maagd et al. (18) may have a radius much larger than those in other studies. Dependence of Surface Adsorption Artifact on Temperature. IGS has been used to measure temperaturedependent Happ for the PAHs (7, 16, 18). For example, Alaee et al. (16) employed IGS to measure Happ of benzene, naphthalene, phenanthrene, and anthracene at different temperatures. For these four aromatic compounds a large number of measurements of vapor pressure and water solubility exist at different temperatures (28). Generally good agreement of the data measured by different investigators gives confidence in their validity, and suggests that it is possible to estimate H as a function of temperature for these four substances (1). Data for P and S from ref 28 were compiled, and best fits for linear regressions between lnP and lnS and reciprocal absolute temperature were obtained. The temperature-dependent H for benzene and the three PAHs is calculated from these regressions and is plotted as a function of temperature in Figure 5. Also included are the experimentally determined Happ for the aromatic compounds by Alaee et al. (16) as well as the Happ estimated using eqs 5 and 6 and different ratios of transfer factor and bubble radius, fTGB/r. For benzene and naphthalene, the H estimated from P and S data agrees very well with the experimental values by Alaee et al. (16) at all temperatures (Figure 5). We also would not expect any difference because the KIA of these two compounds is too small for the surface adsorption artifact to be relevant, irrespective of temperature (Figure 2). Phenanthrene and anthracene, on the other hand, display a discrepancy between calculated H and measured Happ (Figure 5). At lower temperatures the measured H for phenanthrene and anthracene is 2-4 times higher than the P/S ratios, whereas at higher temperature, Happ and H are closer to each other. This temperature dependence of the occurrence of a surface adsorption artifact for these two PAHs is consistent with the magnitude of their predicted KIA value (Figure 2) and the predictions based on eq 6 (Figure 5). Similar to the aromatic compounds, a temperature-dependent surface adsorption artifact was observed for some of the IGS measurements of the H of γ-HCH (refs 12, 15; see Supporting Information).

FIGURE 5. Comparison between Happ for benzene, naphthalene, phenanthrene, and anthracene measured by Alaee et al. (16) (0) and calculated from temperature-dependent vapor pressure and water solubility data (thick straight line, ref 28). The thin curved lines correspond to Happ calculated using different fTGB/r ratios. Figure 5 includes lines corresponding to Happ estimated using eqs 5 and 6. The best fit between measured and predicted Happ is achieved with an fTGB/r in the range of 0.050.15 mm-1 for phenanthrene and anthracene (Figure 5). The bubble radius in the IGS experiment by the investigator is not known. Assuming a bubble radius of 1 mm and also that the estimated KIA for phenanthrene and anthracene is correct, this implies that between 5 and 15% of the chemical adsorbed to the bubble surface is transferred to the gas phase upon bursting. Obviously, larger bubble radii would require larger fTGB values. A consequence of the temperature dependence of the surface sorption artifact is that IGS measurements may yield enthalpies of air-water phase transfer that are too small, i.e., the slope of their log Happ vs 1/T relationship is shallower than it should be (26). The enthalpies of air-water phase transfer obtained by Alaee et al. (16) for phenanthrene and anthracene using the IGS are 32 and 29 kJ‚mol-1. These values are about two times lower than those obtained from the P/S ratios, which are 57 and 54 kJ‚mol-1 for phenanthrene and anthracene, respectively. We also found that IGS measurements by some groups yield lower enthalpies of air-water phase transfer for γ-HCH (see Supporting Information) Flawed Estimation of KIA for Organochlorine Pesticides. Figure 6 compares the Happ for aldrin measured by Cetin et al. (15) with the H calculated using temperature-dependent vapor pressure and solubility data from the literature (refs 108, 36, 86, 110, and 89 from the compilation by Shen and Wania (32)). The figure again includes Happ calculated using eqs 5 and 6, estimated KIA values and different fTGB/r. The best fit between measured and calculated Happ is obtained when the fTGB/r is in the range of 0.00008-0.0004 mm-1. These fTGB/r ratios are unreasonably low and suggests that the KIA for aldrin estimated with eq 7 are much too high. Another indication that the estimated KIA values are too high is the difference in the observed and predicted temperature dependence of the deviation between H and Happ: The measured Happ shows a steady increase in the deviation from H with decreasing temperature, whereas the deviation of the

FIGURE 6. Comparison between Happ of aldrin measured by IGS by Cetin et al. (15) (0) and H calculated from temperature-dependent vapor pressure and water solubility data (thick straight line, ref 32). The measured Happ deviates from the calculated H at lower temperatures. Estimated Happ-values (thin curved lines) yield unreasonably low fTGB/r ratios and a different temperature dependence of the discrepancy between H and Happ than is observed. predicted Happ from H is small at high temperature but very large at low temperatures (Figure 6). The solute descriptors A and B for aldrin are estimated values, and likely fairly uncertain. Also, as discussed above, eq 7 may not be applicable for estimating the KIA of more complex molecules. Not all of the functional groups of pesticide chemicals may be able to make contact with functional groups on the water surface due to their three-dimensional structure (39). We thus conclude that the KIA values estimated for aldrin are considerably too high. This applies presumably also to other pesticide chemicals (e.g., heptachlor, see Supporting Information), preventing a quantitative interpretation of the bias in Happ. Lack of Artifact in Some IGS Studies. The IGS technique has also been used to determine Happ of larger OCPs (13-15) and PBDEs (9, 20). To avoid repetition, the detailed evaluation of these data is relegated to the Supporting Information. This re-evaluation provides somewhat less insight because the VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3811

KIA values estimated for most of these chemicals have to be considered suspect and may be too high by many orders of magnitude. In brief, the Happ values for the OCPs measured by Jantunen and Bidleman (14) and Warner et al. (40) (e.g., heptachlor and dieldrin) and those for the PBDEs by Lau et al. (9) are again too high when compared to final adjusted H-values for the OCPs (32) and PBDEs (30, 31). In contrast to the Happ reported in refs 9, 14, and 40, the Happ measured by Cetin et al. (15) for many OCPs (e.g., heptachlor, aldrin, dieldrin) and by Cetin and Odabasi (20) for the PBDEs are close to the final adjusted H. In contrast to the ten Hulscher et al. (7, 11, 18) studies above, it is not possible to explain this by a higher flow rate of the stripping gas, which would result in larger bubbles. In fact, the flow rate used by Cetin et al. (15, 20) is similar to that used in other IGS studies. Even if the estimated KIA values for PBDEs and OCPs are highly uncertain, we would expect these molecules to adsorb sufficiently strongly to the water surface to be susceptible to the artifact in IGS determinations of H. We have therefore no explanation for why the IGS studies by Cetin et al. (15, 20) appear to give unbiased results for H that are in agreement with other thermodynamic distribution properties for the OCPs and PBDEs. Cetin et al. (15, 20) may be using a method for introducing gas to the stripping vessel that yields very large bubble radii, or otherwise employ experimental conditions that minimize or even eliminate the artifact caused by adsorption to the bubble surface. Mechanism of the Surface Adsorption Artifact. While presenting evidence that organic chemicals that sorb strongly to the air-water interface are often lost from the aqueous solution in IGS experiments beyond what would be expected based on bulk-phase partitioning between air and water, we have so far not yet discussed the mechanism by which such chemicals may be transferred to the gas phase during IGS. We suggest that during the bursting of the gas bubbles at the top of the stripping vessel, chemical is transferred to the head space both as a gas and with newly formed water droplets. In Lei et al. (26) we wrote: “Upon bursting of a gas bubble the surface of that bubble disappears very rapidly, forcing the chemical adsorbed to the bubble surface to either redissolve in the aqueous phase or to enter the gas phase. At the moment of bubble bursting, the fugacity on the rapidly shrinking surface is expected to spike, providing a driving force for chemical transfer in vapor form.” At the same time, during the bubble burst, an aqueous aerosol is being formed, whose surface is loaded with chemical with high KIA. We do not know which of these processes is more important, but expect this to depend on the number and size of the droplets being formed, which in turn will depend on the flow rate and possibly other experimental factors. Using an aerodynamic particle sizer (model 3320, TSI Inc., St. Paul, MN) and our IGS apparatus (2, 26), we have measured aerosol number concentrations between 0.2 and 2 cm-3 and median aerodynamic aerosol diameters between 0.5 and 1.5 µm. The number concentrations increased with gas flow rate. They also were higher when an aqueous solution of normal alkanols (26) was stripped instead of pure water. The formation of aerosol and its potential impact on IGS results is quite well-known (43). As a result, some IGS studies, in particular those that rely on the measurement of vapor concentrations in the stripping gas, have employed aerosol traps (mist traps (9), glass impactors (8), or glass wool plug (13, 15)) to avoid the sampling of droplets in addition to the vapor. IGS studies that rely on the measurement of water concentrations presumably would have to return the aerosol water, and the chemical associated with it, to the stripping vessel, in order not to overestimate the loss of chemical during stripping. Only Lau et al. (9) reported doing this. However, we are suggesting that an aerosol trap is ineffective in achieving either goal. As the water droplets are coalescing 3812

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007

in the aerosol trap, their large surface area vanishes rapidly, resulting in a second fugacity spike, which is expected to drive the chemical to a large extent into the gas phase. In other words, vapor traps will still sample the chemical previously associated with the droplet surface, even if the aerosol itself is not sampled. Similarly, even if the aerosol water is returned quantitatively to the stripping vessel, the chemical from the droplet surface is not. We note that there appears to be no relationship between the use or the type of an aerosol trap in an IGS experiment and the observed size of the surface adsorption artifact. For example, ten Hulscher et al. (11) do not mention an aerosol trap, but Bamford et al. (8) passed the exit gas through a “glass impactor to remove possible aerosols created by breaking bubbles”. The differences between different IGS studies in terms of the occurrence and size of the surface adsorption artifact can thus not be explained by the extent to which the formation of aerosols was taken into consideration. Suitability of IGS for Measuring H of Semivolatile Organic Compounds. Overall, we conclude that the deviations between the Happ measured by IGS and the H judged correct based on other physical-chemical properties (i.e., based on P/S ratios or FAVs), when indeed observed, (i) are almost always of the direction that would be expected based on a surface adsorption artifact, namely the H from IGS is generally too high, (ii) within a group of SOCs (e.g., within the group of PAHs and PCBs) generally show the dependence with compound size that would be expected based on a surface sorption artifact, namely the discrepancy increases with increasing size and therefore KIA, (iii) always show the dependence on temperature that would be expected based on a surface adsorption artifact, namely the discrepancy increases with decreasing temperature, and (iv) within the considerable uncertainty of the estimate of KIA, fTGB, and r are generally of a magnitude that would be expected based on a surface sorption artifact and eqs 5 and 6. We thus believe that there is sufficient evidence to conclude that IGS often provides unreliable results for SOCs larger than HCB, HCHs, and naphthalene, especially at lower temperatures, because of adsorption to the bubble surface and transfer to the gas phase upon bursting (26). Many of the Happ-values reported for SOCs have to be considered biased high, and the reported temperature dependence of the air-water partitioning, i.e., the enthalpy or internal energy of air-water phase transfer, of those SOCs is too small. However, our data analysis also reveleaed that some investigators (7, 11, 15, 18, 20) have used IGS to generate H-values for SOCs which are close to the H values judged to be correct. Different measures to prevent the sampling of aerosol formed during bubble bursting cannot explain why some IGS studies succeed and other do not. Our hypothesis is that successful IGS studies used larger bubble sizes, and the use of IGS apparati that generate large bubbles may indeed significantly reduce the size of a surface adsorption artifact. However, when using bigger bubble radii, it still has to be assured that equilibrium between dissolved and gas phase is achieved. Whereas the IGS method can provide reliable information on the air-water partitioning of substances with a KIA below 0.001 m, it clearly cannot be recommended for determining H for highly surface active chemicals, until it is more clearly understood what experimental conditions help to avoid a surface adsorption artifact.

Acknowledgments We are grateful for financial support from the Natural Sciences and Engineering Research Council of Canada and for the assistance of Jay Slowik with the aerosol measurements.

Supporting Information Available Tables with solute descriptors for the SOCs; text and figures with evaluation of additional IGS-derived Happ data for OCPs and PBDEs. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry; John Wiley & Sons: New York, 1993. (2) Mackay, D.; Shiu, W.-Y.; Sutherland, R. P. Determination of air-water Henry’s law constants for hydrophobic pollutants. Environ. Sci. Technol. 1979, 13, 333-337. (3) Bruhn, R.; Lakaschus, S.; McLachlan, M. S. Air/sea gas exchange of PCBs in the Southern Baltic Sea. Atmos. Environ. 2003, 37, 3445-3454. (4) Li, Y.-F.; MacDonald, R. W.; Jantunen, L. M. M.; Harner, T.; Bidleman, T. F.; Strachan, W. M. J. The transport of β-hexachlorocyclohexane to the western Arctic Ocean: a contrast to R-HCH. Sci. Total Environ. 2002, 291, 229-246. (5) Leroi, J.-C.; Masson, J.-C.; Renon, H.; Fabries, J.-F.; Sannier, H. Accurate measurements of activity coefficients at infinite dilution by inert gas stripping and gas chromatography. Ind. Eng. Chem. Process Des. Dev. 1977, 16, 139-144. (6) Sashuvar, L.; Helm, P. A.; Jantunen, L. M. M.; Bidleman, T. F. Henry’s law constants for R-, β-, and γ-hexachlorocyclohexanes (HCHs) as a function of temperatures and revised estimates of gas exchange in Arctic regions. Atmos. Environ. 2003, 37, 983992. (7) ten Hulscher, Th. E. M.; van der Velde, L. E.; Bruggeman, W. A. Temperature dependence of Henry’s law constants for selected chlorobenzenes, polychlorinated biphenyls and polycyclic aromatic hydrocarbons. Environ. Toxicol. Chem. 1992, 11, 15951603. (8) Bamford, H. A.; Poster, D. L.; Baker, J. E. Henry’s law constants of polychlorinated biphenyl congeners and their variation with temperature. J. Chem. Eng. Data 2000, 45, 1069-1074. (9) Lau, F. K.; Charles, M. J.; Cahill, T. M. Evaluation of gas stripping methods for the determination of Henry’s law constants for polybrominated diphenyl ethers and polychlorinated biphenlys. J. Chem. Eng. Data 2006, 51, 871-878. (10) Fang, F.; Chu, S.; Hong, C.-S. Air-water Henry’s law constants for PCB congeners: Experimental determination and modeling of structure-property relationship. Anal. Chem. 2006, 78, 54125418. (11) ten Hulscher, Th. E. M.; van den Heuvel, H.; van Noort, P. C. M.; Govers, H. A. J. Henry’s law constants for eleven polychlorinated biphenyls at 20 °C. J. Chem. Eng. Data 2006, 51, 347 -351. (12) Kucklick, J. R.; Hinckley, D. A.; Bidleman, T. F. Determination of Henry’s law constants for hexachlorocyclohexanes in distilled water and artificial seawater as a function of temperature. Mar. Chem. 1991, 34, 197-209. (13) Jantunen, L. M. M.; Bidleman, T. F. Temperature dependent Henry’s law constant for technical toxaphene. Chemosphere: Global Change Sci. 2000, 2, 225-231. (14) Jantunen, L. M. M.; Bidleman, T. F. Henry’s law constants for hexachlorobenzene, p,p′-DDE and components of technical chlordane and estimates of gas exchange for Lake Ontario. Chemosphere 2006, 62, 1689-1696. (15) Cetin, B.; Ozer, S.; Sofuoglu, A.; Odabasi, M. Determination of Henry’s law constants of organochlorine pesticides in deionized and saline water as a function of temperature. Atmos. Environ. 2006, 40, 4538-4546. (16) Alaee, M.; Whittal, R. M.; Strachan, W. M. J. The effect of water temperature and composition on Henry’s law constant for various PAHs. Chemosphere 1996, 32, 1153-1164. (17) Shiu, W.-Y.; Mackay, D. Henry’s law constants of selected aromatic hydrocarbons, alcohols and ketones. J. Chem. Eng. Data 1997, 42, 27-30. (18) De Maagd, P. G.-J.; ten Hulscher, Th. E. M.; van den Heuvel, H.; Opperhuizen, A.; Sijm, D. T. H. M. Physicochemical properties of polycyclic aromatic hydrocarbons aqueous solubilities, n-octanol/water partition coefficients and Henry’s law constants. Environ. Toxicol. Chem. 1998, 17, 251-257. (19) Bamford, H. A.; Poster, D. L.; Baker, J. E. Temperature dependence of Henry’s law constants of thirteen polycyclic aromatic hydrocarbons between 4 °C and 31 °C. Environ. Toxicol. Chem.1999, 18, 1905-1912.

(20) Cetin, B.; Odabasi, M. Measurement of Henry’s law constants of seven polybrominated diphenly ether (PBDE) congeners as a function of temperature. Atmos. Environ. 2005, 39, 52735280. (21) Goss, K.-U.; Wania, F.; McLachlan, M. S.; Mackay, D.; Schwarzenbach, R. P. Comment on “Re-evaluation of air-water exchange fluxes of PCBs in Green Bay and Southern Lake Michigan”. Environ. Sci. Technol. 2004, 38, 1626-1628. (22) Brunner, S.; Hornung, E.; Santl, H.; Wolff, E.; Piringer, O. G.; Altschuh, J.; Bruggemann, R. Henry’s law constants for polychlorinated biphenyls: experimental determination and structure-property relationships. Environ. Sci. Technol. 1990, 24, 1751-1754. (23) Murphy, T. J.; Mullin, M. D.; Meyer, J. Equilibrium of polychlorinated biphenyls and toxaphene with air and water. Environ. Sci. Technol. 1987, 21, 155-162. (24) Li, N.; Wania, F.; Lei, Y. D.; Daly, G. L. A comprehensive and critical compilation, evaluation and selection of physicalchemical property data for selected polychlorinated biphenyls. J. Phys. Chem. Ref. Data 2003, 32, 1545-1590. (25) Schenker, U.; Macleod, M.; Scheringer, M.; Hungerbu ¨ hler, K. Improving data quality for environmental fate models: A leastsquares adjustment procedure for harmonizing physicochemical properties of organic compounds. Environ. Sci. Technol. 2005, 39, 8434-8441. (26) Lei, Y. D.; Shunthirasingham, C.; Wania, F. Comparison of headspace and gas-stripping techniques for measuring the airwater partitioning of normal alkanols (C4 to C10) - effect of temperature, chain length and adsorption to the water surface. J. Chem. Eng. Data 2007, 52, 168-179. (27) Roth, C. M.; Goss, K.-U.; Schwarzenbach, R. P. Adsorption of a diverse set of organic vapors on the bulk water surface. J. Colloid Interface Sci. 2002, 252, 21-30. (28) Mackay, D.; Shiu, W. Y.; Ma, K. C.; Lee, S. C. Handbook of PhysicalChemical Properties and Environmental Fate for Organic Chemicals. Vol. III Oxygen Containing Compounds, 2nd ed.; CRC Press, Taylor & Francis Group: Boca Raton, FL, 2006. (29) Xiao, H.; Li, N.; Wania, F. Compilation, evaluation and selection of physical-chemical property data for R-, β-, and γ-hexachlorocyclohexane. J. Chem. Eng. Data 2004, 49, 173-185. (30) Wania, F.; Dugani, C. B. Assessing the long-range transport potential of polybrominated diphenyl ethers: A comparison of four multimedia models. Environ. Toxicol. Chem. 2003, 22, 1252-1261. (31) Palm, A.; Brorstro¨m-Lunde´n, E.; Breivik, K. Transport and fate of polybrominated diphenyl ethers in the Baltic and Arctic regions; Report to the Nordic Council of Ministers; TemaNord 2004:554; Norden: Copenhagen, 2004; Appendix 1, pp 65-75. (32) Shen, L.; Wania, F. Compilation, evaluation and selection of physical-chemical property data for organochlorine pesticides. J. Chem. Eng. Data 2005, 50, 742-768. (33) Lei. Y. D.; Wania, F. Is rain or snow a more efficient scavenger of organic chemicals? Atmos. Environ. 2004, 38, 3557-3571. (34) Goss, K.-U.; Predicting the equilibrium partitioning of organic compounds using just one linear solvation energy relationship (LSER). Fluid Phase Equilib. 2005, 233, 19-22. (35) Abraham, M. H.; Enomoto, K.; Clarke, E. D.; Sexton, G. Hydrogen bond basicity of the chlorogroup; hexachlorocyclohexanes as strong hydrogen bond bases. J. Org. Chem. 2002, 67, 47824786. (36) Abraham, M. H.; Al-Hussaini, A. J. M. Solvation parameters for the 209 PCBs: calculation of physicochemical properties. J. Environ. Monit. 2006, 7, 295-301. (37) Platts, J. A.; Butina, D.; Abraham, M. H.; Hersey, A. Estimation of molecular liner free energy relation descriptors using a group contribution approach. J. Chem. Info. Comput. Sci. 1999, 39, 835-845. (38) Roth, C. M.; Goss, K.-U.; Schwarzenbach, R. P. Sorption of diverse organic vapors to snow. Environ. Sci. Technol. 2004, 38, 40784084. (39) Goss, K.-U. The air/surface adsorption equilibrium of organic compounds under ambient conditions. Crit. Rev. Environ. Sci. Technol. 2004, 34, 339-389. (40) Warner, H. P.; Cohen, J. M.; Ireland, J. C. Determination of Henry’s law constants of selected priority pollutants; EPA/600/D-87/ 229; NTIS PB87-212684; U.S. Environmental Protection Agency: Cincinnati, OH, 1987. (41) Hinckley, D. A.; Bidleman, T. F.; Foreman, W. T.; Tuschall, J. R. Determination of vapor pressures for nonpolar and semipolar VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3813

organic compounds from gas chromatographic retention data. J. Chem. Eng. Data 1990, 35, 232-237. (42) Biggar, J. W.; Riggs, R. L. Apparent solubility of organochlorine insecticides in water at various temperatures. Hilgardia 1974, 42, 383-391. (43) Friesen, K. J.; Fairchild, W. L.; Loewen, M. D.; Lawrence, S. G.; Holoka, M. H.; Muir, D. C. G. Evidence for particle-mediated

3814

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007

transport of 2,3,7,8-tetrachlorodibenzofuran during gas sparging of natural waters. Environ. Toxicol. Chem. 1993, 12, 2037-2044.

Received for review December 13, 2006. Revised manuscript received March 23, 2007. Accepted March 28, 2007. ES062957T