Environ. Sci. Technol. 2005, 39, 8840-8846
Improved Prediction of Octanol-Water Partition Coefficients from Liquid-Solute Water Solubilities and Molar Volumes C A R Y T . C H I O U , * ,† DAVID W. SCHMEDDING,‡ AND MILTON MANES§ U.S. Geological Survey, Box 25046, Mail Stop 408, Federal Center, Denver, Colorado 80225, Department of Environmental and Molecular Toxicology, Oregon State University, Corvallis, Oregon 97331, and Department of Chemistry, Kent State University, Kent, Ohio 44242
A volume-fraction-based solvent-water partition model for dilute solutes, in which the partition coefficient shows a dependence on solute molar volume (V), is adapted to predict the octanol-water partition coefficient (Kow) from the liquid or supercooled-liquid solute water solubility (Sw), or vice versa. The established correlation is tested for a wide range of industrial compounds and pesticides (e.g., halogenated aliphatic hydrocarbons, alkylbenzenes, halogenated benzenes, ethers, esters, PAHs, PCBs, organochlorines, organophosphates, carbamates, and amidesureas-triazines), which comprise a total of 215 test compounds spanning about 10 orders of magnitude in Sw and 8.5 orders of magnitude in Kow. Except for phenols and alcohols, which require special considerations of the Kow data, the correlation predicts the Kow within 0.1 log units for most compounds, much independent of the compound type or the magnitude in Kow. With reliable Sw and V data for compounds of interest, the correlation provides an effective means for either predicting the unavailable log Kow values or verifying the reliability of the reported log Kow data.
Introduction Octanol-water partition coefficient (Kow) and water solubility are two fundamental descriptors used to assess the transport and fate of organic compounds in environmental systems (1, 2). Kow serves not only as a general indicator for a compound’s tendency to partition into an organic phase, this coefficient is practically the same as the compound’s lipid (triolein)-water partition coefficient (Ktw), the latter accounting directly for the fish bioconcentration factor on a lipid-weight basis (3). Despite the usefulness of Kow in environmental research, accurate Kow values remain unavailable for numerous compounds. On the other hand, for many compounds with available Kow data, the reported Kow * Corresponding author phone: (303) 236-3967; fax: (303) 2363934; e-mail:
[email protected]. † U.S. Geological Survey. ‡ Oregon State University. Current Address: 29479 Beaver Creek Road, Corvallis, Oregon 97333. § Kent State University. Current Address: Amberson Towers #412, 5 Bayard Road, Pittsburgh, Pennsylvania 15213. 8840
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
values for a given compound are often vastly inconsistent. Indeed, it is not uncommon to see the discrepancy in reported Kow for a compound from different laboratories or by different analytical methods exceed 1-2 orders of magnitude. Such a wide spread in Kow poses an immense burden on the user to select reliable Kow values for the transport and fate assessment. Water solubility (Sw) of a partially soluble liquid or supercooled liquid may be thought of as a special form of the solute partition coefficient, namely that between its own excess liquid phase and water (4); the supercooled-liquid Sw of a solid may be obtained via a thermodynamic conversion method from the solid Sw, to be addressed later. Therefore, the liquid Sw is also a good general indicator of a compound’s partition tendency with an organic medium. As noted, the magnitude of Kow depends largely on liquid Sw (5-7). Compared to Kow, the Sw data for conventional chemicals are usually more available and accurate because of traditional interests in their solution properties. Nonetheless, for sparingly soluble compounds, such as polychlorinated biphenyls (PCBs), polycyclic aromatic hydrocarbons (PAHs), chlorinated dioxins, phthalates, and many pesticides, accurate Sw data remain sparse. Similar to the case with Kow, available Sw data for given low-solubility compounds spread widely, subject to further resolution. Generally, accurate Kow and Sw values are obtained with careful steps to minimize measurement errors, such as those caused by the impurities of test compounds and solvents, the equilibration and separation methods, and the equipment sensitivity for detecting target compounds (4). Determination of accurate Kow and Sw for highly insoluble compounds becomes challenging because of the limitation in detecting trace solutes in water and the stringent effects of emulsions and phase separation on solute concentrations. Currently, the most consistent, and seemingly most reliable, analytical methods for Kow and Sw are the shake-flask (usually coupled with centrifugation) (4, 8) and the generator-column equilibration techniques (9, 10). Indirect experimental methods, e.g., those by HPLC retention times (11, 12), and empirical computation models, e.g., those by fragment constants (13, 14), have been developed for predicting Kow and/or Sw but their uses are usually confined to relatively simple molecules or those within a homologous series (15). Correlations between log Kow and log Sw have been reported (5-7). Such correlations, however, usually do not cover a wide diversity of compounds and a wide range of Kow and Sw with the same accuracy for resolving the inconsistency of multiple Kow and Sw values of given compounds. Presently, accurate estimation of log Kow, with a standard deviation (SD) ) 0.12, could be achieved via polyparameter linearsolvation-energy relationships (pp-LSERs) (16, 17); the needed parameter values, however, are not available for many environmental compounds, especially the pesticides (17, 18). We here present a significantly improved log Kow-log Sw correlation that takes into account the solute-size effect on the partition coefficient. This new correlation is tested for a wide variety of industrial chemicals and pesticides that span some 8.5 orders of magnitude in Kow and 10 orders of magnitude in Sw. As illustrated later, it enables one to predict log Kow from reliable log Sw, or vice versa, with exceptional accuracies and thus may be used as a rapid tool to identify credible Kow and/or Sw data.
Theory The partition coefficients for (dilute) solutes of limited water solubility in a two-phase solvent-water mixture are governed 10.1021/es050729d CCC: $30.25
2005 American Chemical Society Published on Web 10/08/2005
by their solubilities in water and solvent phases. The solute compatibility with a solvent is measured by the proximity of the actual solution to the perfect-solution state. Conventionally, the perfect solution for a liquid solute is defined as one in which the solute activity equals its mole fraction (19). The solute activity may also be expressed in terms of the volume fraction (9, 20). The perfect-solution model is particularly adaptable for liquids completely miscible with the solvent. The relative merits of the two solution models to account for the solute partition behavior become evident later when they are applied to analyze the octanol-water partition coefficients of a wide variety of compounds. Using the mole fraction as a basis to express the solute activity, Chiou et al. (7) derived the octanol-water partition coefficient (Kow) for partially water-soluble organic solutes as
log Kow ) -log Sw - log
V/o
- log
γ/o
- log
(γw/γ/w)
(1)
where Sw is the molar solubility of a liquid or supercooledliquid solute in water (mol/L); V/o is the molar volume (L/ mol) of the water-saturated octanol, 0.120 L/mol; γ/o is the mole-fraction-based solute activity coefficient in watersaturated octanol (dimensionless); γw is the solute activity coefficient in pure water; and γ/w is the similar activity coefficient in octanol-saturated water. The log(γw/γ/w) expresses the solute solubility enhancement in water by dissolved octanol. The sum of log γ/o and log(γw/γ/w) is the total deviation of the experimental log Kow from the theoretical partition coefficient (log Koow) where the solute forms a perfect solution in water-saturated octanol (i.e., γ/o ) 1) and where the dissolved octanol in water has no effect on solute solubility (i.e., γw/γ/w ) 1). Thus, by this solution model, one obtains a hypothetical perfect octanol-water partition coefficient (Koow) for a dilute partially soluble solute as
log Koow ) - log Sw - log V/o
(2)
As noted with eqs 1 and 2, the partition coefficient as derived shows a specific dependence on the solvent molar volume (V/o). Analyses of the Kow and Sw data for 36 substituted aromatic solutes (primarily substituted benzenes) with eq 1 indicate that the most influential factor on Kow is Sw, followed by γ/o and γw/γ/w, respectively (7). Both log γ/o and log(γw/γ/w) are found to increase largely in proportion with decreasing log Sw. For these solutes, regression of the log Kow against the respective log Sw gives
log Kow ) -0.862 log Sw + 0.710
(3)
with R2 ) 0.990. As described later, whereas eq 3 predicts well the log Kow of solutes whose molar volumes (V) are comparable to those of substituted benzenes, it tends to underestimate the log Kow for small-sized solutes and overestimate the log Kow for large-sized solutes to varying extents. The apparent dependence of Kow on solute size is not predicted by the preceding model, which predicts instead a dependence on the solvent size (eqs 1 and 2) that arises from the use of mole fraction to express the solute activity at dilution. The presumed solute-size effect on Kow appears to be a major reason that precludes accurate estimation of log Kow for different-sized solutes from their log Sw or other molecular properties. We now consider the alternative log Kow derivation using the volume fraction to express the solute activity. On this basis, Tewari et al. (9) and Sangster (20) derived an expression for Kow while giving no explicit accounts of the effects of Sw and octanol-water mutual saturation on Kow. Accounting
all the factors as with eq 1, one arrives at the expression for log Kow as
log Kow ) -log Sw - log V - log Fdv
(4)
where V is the solute molar volume and log Fdv ) log F1 + log F2, with log F1 ) log γ/o and log F2 ) log(γw/γ/w). Here γ/o, γw, and γ/w are the activity coefficients, analogous to those in eq 1, on a volume-fraction basis. In eq 4, the solubility of water in octanol (ca. 5 wt %) is assumed to have a negligible effect on the octanol solvency (alternatively, it may be treated as added log F3 to log Fdv). The perfect partition coefficient for a partially water-soluble solute (Kosw), which is the same in any two-phase solvent-water mixture, may thus be given as
log Kosw ) -log Sw - log V
(5)
The implied independence of Kow to solvent size (V/o) in eqs 4 and 5 is supported by the observation that the log Kow for various solutes are similar in magnitude to the respective log Ktw (triolein-water), despite the fact that triolein is 6 times as large as octanol (3). The implied dependence of Kow on solute V may be rationalized in terms of the mixing entropy as follows. When two liquid solutes of unequal sizes with identical molar Sw partition from water into a fixed volume of the organic-solvent phase, the smaller solute will undergo more volume expansion to score more gain in entropy than the larger solute. If the two solutes exhibit similar thermal effects with the solvent, the smaller solute will then have a lower chemical potential (or activity) to favor its partition. The accuracy in predicting log Kow for various solutes depends on the accuracy in predicting their log Fdv from the respective log Sw (or other properties). Lumping log F1 and log F2 into a single log Fdv facilitates estimation of the total deviation of log Kow from log Kosw, since both the (molefraction-based) log γ/o and log(γw/γ/w) terms in eq 1 are known to be inversely related to log Sw (7). To achieve the objective, we first calculate, by eqs 4 and 5, the individual log Fdv values for selected reference solutes that have relatively accurate log Kow, log Sw, and log V data. The resulting log Fdv values are then regressed against the respective log Sw to establish the correlation between log Fdv and log Sw for the subsequent log Kow prediction with other compounds. The predictive power of the model is critically examined with more than 200 compounds from various chemical classes.
Data Selection and Handling The various reported Sw and log Kow values for given compounds are examined for reliability and consistency. Only the direct experimental Sw and log Kow values are evaluated, since accuracies of the data from indirect experimental or other calculation methods are difficult to assess; further, only the data at 25 °C or otherwise at room temperature (20-30 °C) are considered. Compilations of reported Sw and log Kow data and their analytical methods for a large number of environmental compounds are available from Mackay et al. (21-25). For compounds with multiple reported Sw and log Kow values, the data are first screened on the basis of the methods used and the mutual consistency from independent sources. The data generated by shake-flask or generatorcolumn equilibration with identified analytical methods (e.g., GC, UV, and HPLC) are given a priority for evaluation and selection. Reported data without identified methods are used only when the aforementioned data are not available. When widespread Sw values exist for a compound, the one that yields a predicted log Kow in best agreement with the reported log Kow is considered most reliable and selected. This acceptance is based on the observed close fit of the predicted VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8841
TABLE 1. Log Sw, log Kow, log V, log Fdv, and Other Properties of Selected Reference Compounds at 25 °C compound diethyl ether di-n-propyl ether aniline 2-toluidine 3-chloroaniline dichloromethane chloroform carbon tetrachloride benzene toluene 1,2-xylene ethylbenzene chlorobenzene bromobenzene 1,3-dichlorobenzene 1,2,3-trichlorobenzene 1,2,3,4-tetrachlorobenzene 1,2,4,5-tetrachlorobenzene n-hexane n-octane 1-hexene biphenyl naphthalene 2,6-dimethylnaphthalene phenanthrene 3-PCB 2,4,5-PCB 2,2′,5-PCB 2,2′,4,5,5′-PCB 2,2′,3,3′,4,4′-PCB chlorpyrifos lindane p,p′-DDT
M (g/mol)
mp (°C)a
74.1 102.2 93.1 107.2 127.6 84.9 119.4 153.8 78.1 92.1 106.2 106.2 112.6 157.0 147.0 181.5 215.9 215.9 86.2 114.2 84.2 154.2 128.2 156.2 178.2 188.7 257.5 257.5 326.4 360.9 350.6 290.9 354.5
Lc L L L L L L L L L L L L L L 53 48 140 L L L 71 80 110 101 L 79 44 77 150 43 113 109
-log V (L/mol) 0.984 0.861 1.04 0.968 0.979 1.19 1.09 1.01 1.05 0.974 0.918 0.912 0.993 0.979 0.943 0.903 0.848 0.848 0.881 0.787 0.900 0.749 0.903 0.818 0.773 0.717 0.649 0.649 0.593 0.526 0.601 0.762 0.603
Sw (mg/L)
ref
6.03E4 2.5E3 3.62E4 1.63E4 5.44E3 1.94E4 7.23E3 800 1.78E3 517 175 159 472 410 124.5 16.3 3.42 0.29 9.50 0.66 69.7 6.71 31.7 0.997 1.29 3.63 0.092 0.248 1.03E-2 2.85E-4 0.40 7.87 5.5E-3
27 29 30 7 7 34 35 34 36 37 38 36 41 42 10 3 43 43 36 36 40 10 44 46 47 48 50 8 8 10 4 52 50
∆Hf (cal/mol)a NAd NA NA NA NA NA NA NA NA NA NA NA NA NA NA 4150 4060 5760 NA NA NA 4180 4540 5800 4450 NA 5450 4280 4490 6980 6200 5640 6300
-log Sw (mol/L)b
log Kosw
log Kow
ref
log Fdv
0.0899 1.61 0.410 0.817 1.37 0.641 1.22 2.28 1.64 2.25 2.78 2.82 2.38 2.58 3.07 (3.79) (4.59) (4.70) 3.69 5.24 3.08 (3.95) (3.09) (4.25) (4.48) 4.72 (5.83) (5.83) (7.01) (7.59) (5.68) (3.62) (6.79)
1.07 2.47 1.45 1.78 2.35 1.83 2.31 3.30 2.69 3.22 3.70 3.74 3.37 3.56 4.01 4.69 5.43 5.54 4.84 6.02 3.98 4.70 3.99 5.07 5.25 5.43 6.48 6.48 7.61 8.12 6.29 4.39 7.40
0.83 2.03 1.09 1.42 1.90 1.51 1.90 2.73 2.13 2.69 3.13 3.15 2.84 2.99 3.44 4.04 4.60 4.70 4.11 5.18 3.39 4.04 3.36 4.31 4.46 4.58 5.51 5.60 6.50 6.98 5.27 3.72 6.36
28 30 31 32 33 33 35 35 13 13 39 40 13 13 35 10 3 3 40 40 40 35 45 33 33 49 10 49 49 10 51 53 7
0.24 0.44 0.36 0.36 0.45 0.32 0.41 0.57 0.56 0.53 0.57 0.59 0.53 0.57 0.57 0.65 0.83 0.84 0.73 0.84 0.59 0.66 0.63 0.76 0.79 0.85 0.97 0.88 1.11 1.14 1.02 0.67 1.04
a Melting points (mp), molar heats of fusion (∆H ), and densities (for calculation of V) are from those cited in Mackay et al. (21-25). The ∆H f f for 2,2′,5-PCB is estimated with ∆Hf ) 13.5Tm. b Values with parentheses are for the supercooled liquids. c Liquids at 25 °C. d NA ) not applicable for liquids.
log Kow to the well-determined log Kow for a wide range of compounds, to be illustrated. The molar volumes (V) of compounds required for model testing and prediction are obtained from their molecular weights (M) and liquid (or supercooled-liquid) densities (d) at 25 °C or room temperature. For solid compounds, the supercooled-liquid densities are approximated to equal the solid densities if the melting points (mp) are < 100 °C; if the melting points are > 100 °C, the supercooled liquid densities are approximated to be 90% of the solid densities, based on an earlier observation (3), except where the liquid densities have been measured. In a few cases where the densities of compounds are unavailable, they are taken from the values of their liquid isomers or compositionally similar liquids. In general, the solute density (or molar volume) is relatively temperature-independent and can be easily measured in the laboratory. Errors on the estimated liquid densities are judged to be e 10%. For solid compounds at 25 °C, the Sw values for their supercooled liquids are calculated from the corresponding solid solubilities using the thermodynamic conversion factor, Fsl (dimensionless), i.e.,
Sw (liquid) ) FslSw (solid)
(6)
with
log Fsl )
∆Hf(Tm - T) 2.303RTTm
(7)
where ∆Hf is the compound’s molar heat of fusion (cal/mol), 8842
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
Tm is the melting point (K), T is the system temperature (K) taken as 298 K, and R is the gas constant (1.987 cal/K-mol). The ∆Hf values for solid compounds are taken from sources cited in Mackay et al. (21-25). When ∆Hf data are unavailable, approximate values are obtained using ∆Hf ) 13.5Tm, which holds well for many hydrocarbons (5-7, 26). Considering that the ∆Hf/Tm values fall mostly between 9 and 18, the error in log Fsl by using ∆Hf ) 13.5Tm should largely be < 0.20 units for compounds with mp < 100 °C; the error may increase up to 0.40 units for compounds with mp ) 150 °C. In this work, we limit the use of ∆Hf ) 13.5Tm primarily for compounds with mp < 130 °C when experimental ∆Hf values are unavailable.
Results and Discussion The perfect-solution log Kosw and experimental log Kow for 33 selected reference solutes and their respective log Sw and log Fdv values are listed in Table 1. The selection of these solutes is somewhat arbitrary, based primarily on the credibility of analytical methods, the spans of log Sw and log Kow, and the diversity of chemical classes. As mentioned, the log Sw values for solid compounds are those of the supercooled liquids obtained via eqs 6 and 7. By analogy to the observation that both log γ/o and log(γw/γ/w) in eq 1 are inversely related to log Sw (7), it is sensible to correlate log Fdv in eq 4 with log Sw of the solutes. Using the data in Table 1, the regression gives
log Fdv ) -0.116 log Sw + 0.268
(8)
with n ) 33 and R2 ) 0.972. A plot of log Fdv versus -log Sw for reference compounds is shown in Figure 1. As observed,
FIGURE 1. Correlation of log Fdv with -log Sw for reference compounds from Table 1. whereas the Sw extend over nearly 7 orders of magnitude, the log Fdv vary over a relatively small range, from 0.24 for diethyl ether to 1.14 for 2,2′,3,3′,4,4′-PCB. Substituting eq 8 into eq 4, one obtains a specific solute-size-corrected relation between log Kow and log Sw as
log Kow ) -0.884 log Sw - log V - 0.268
(9)
The merit of eq 9 over eq 3 for predicting log Kow is examined with a wide variety of industrial chemicals and pesticides using the available log Sw and log V data. Industrial chemicals and numbers of compounds (n) tested are as follows: aliphatic hydrocarbons (ALHCs) (n ) 14);
halogenated aliphatic hydrocarbons (HALHCs) (n ) 22); alkylbenzenes (ALBZs) (n ) 15); halogenated benzenes (HABZs) (n ) 14); substituted anilines (n ) 6); phenols (n ) 15); ethers (n ) 7); esters (n ) 11); alcohols (n ) 6); heterocyclics (n ) 6); PAHs (n ) 23); PCBs (n ) 26); and dibenzodioxins-dibenzofurans (DDXDFs) (n ) 5). Pesticides tested include the following: organochlorines (OGCLs) (n ) 7); organophosphates (OGPPs) (n ) 14); carbamates (n ) 10); and amides-ureas-triazines (AUTZs) (n ) 14). The entire list gives a total of 215 compounds (i.e., 170 industrial chemicals and 45 pesticides) of varying polarities that span about 10 orders of magnitude in Sw and 8.5 orders of magnitude in Kow. Tables showing log V h , accepted literature log Sw and log Kow, and predicted log Kow for compounds from various classes are presented in the Supporting Information (Tables 1S-17S). As shown in Tables 1S-17S, the log Kow predicted by eq 9 for compounds from all classes, except phenols and alcohols to be addressed later, agree remarkably closely with the experimental log Kow, with the discrepancies being 100 °C, the supercooled liquid density is assumed to be 0.90 times the solid density. The italicized ∆Hf values are estimated with ∆Hf ) 13.5 Tm. b Sw ) FslS/w. c Liquids at 25 °C. d NA ) not applicable for liquids. e 4EOs ) four ethylene-oxide units.
VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8843
heterocyclics (n ) 6). For the pesticides, the deviations are the following: 0.03 for OGCLs (n ) 7); 0.11 for OGPPs (n ) 14); 0.07 for carbamates (n ) 10); and 0.07 for AUTZs (n ) 14). For the total set (n ) 194), absolute ∆ log Kow ) 0.07. By eq 9, the accepted log Sw and log Kow are generally those produced with credible analytical methods. Over more than 8 orders of magnitude in Kow examined, the ∆ log Kow show no obvious bias on the compound type or the magnitude in Kow. From the above findings, the present predictive model using only two parameters predicts log Kow at least as well as the five-parameter pp-LSER (16), which offers one of the most accurate predictions among all estimation methods with a SD ) 0.12 units. Moreover, the current approach provides a consistency check between Sw and Kow, which is not achievable with the pp-LSER model. The sharp log Kow prediction by eq 9 for the compounds examined manifests the following: (i) the predominant effect of -log Sw on log Kow for the compounds (7), (ii) the predicted accuracy in log Fdv for all compounds by taking the solutesize correction, and (iii) the supremacy of the volumefraction-based solution model over the mole-fraction-based model to describe the solute partitioning. In some cases where absolute ∆ log Kow values slightly exceed 0.2, as for ethyliodide, di-i-propyl ether, 4-methylbiphenyl, 4,4′-dimethylbiphenyl, 2,3,4,5-PCB, thiophene, methyl parathion, and fonofos, the existing data on either Sw or log Kow are too limited to closely validate the Sw or log Kow data. For a few compounds with mp > 130 °C without experimental ∆Hf, e.g., benz[a]anthracene (159 °C), benzo[a]pyrene (252.3 °C), 4,4′-PCB (149 °C), 1,2,3,4-tetrachlorodioxin (189 °C), carbofuran (152 °C), atrazine (175 °C), and simazine (227 °C), the log Kow predicted with estimated supercooled-liquid Sw using ∆Hf ) 13.5 Tm agree closely with the experimental log Kow (with absolute ∆ log Kow < 0.2). This observation suggests that ∆Hf ) 13.5 Tm is a good approximation for these compounds. Improvement of the solute-size-corrected model (eq 9) for predicting log Kow is best noticed with small- and largesized compounds. Consider first the small-sized solutes, e.g., chloroform, 1,2-dichloroethane, and trichloroethylene, where V ) 0.079-0.090 L/mol. For these solutes eq 9 well predicts the log Kow, with absolute ∆ log Kow < 0.03, whereas eq 3 underpredicts the log Kow by 0.10-0.14. For large-sized solutes, e.g., ethion, leptophos, 2,2′,3,3′,5,5′,6,6′-PCB, and 2,2′,3,3′,4,5,5′,6,6′-PCB, with V ) 0.27-0.34 L/mol, eq 9 predicts the log Kow within 0.06, whereas eq 3 overpredicts by 0.29-0.42. In contrast, for simple substituted benzenes, e.g., toluene, 1,2-dichlorobenzene, and 1,2,4-trichlorobenzene with intermediate V ) 0.10-0.14 L/mol, the log Kow predicted with eq 3 and eq 9 agree closely, differing by 105. Ecotoxicol. Environ. Saf. 1986, 11, 251-260. (61) Bowman, B. T.; Sans, W. W. Determination of octanol-water partitioning coefficients (Kow) of 61 organophosphorous and carbamate insecticides and their relationship to respective water solubility (S) values. J. Environ. Sci. Health 1983, B18 (6), 667683. (62) Briggs, G. G. Theoretical and experimental relationships between soil adsorption, octanol-water partition coefficients, water solubilities, bioconcentration factors and Parachlor. J. Agric. Food Chem. 1981, 29, 1050-1059. (63) Hartley, G. S.; Graham-Bryce, I. J. Physical Principles of Pesticide Behaviour; Academic Press: New York, 1980. (64) Leo, A.; Hansch, C.; Elkins, D. Partition coefficients and their uses. Chem. Rev. 1971, 71, 525-616.
8846
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005
(65) Woodford, E. K., Evans, S. A., Eds. Weed Control Handbook: Properties of Herbicides; Blackwell Scientific: Oxford, England, 1963. (66) Varhanickova, D.; Shiu, W. Y.; Mackay, D. Aqueous solubilities of alkylphenols and methoxyphenols at 25 °C. J. Chem. Eng. Data 1995, 40, 448-451. (67) Wasik, S. P.; Tewari, Y. B.; Miller, M. M.; Martire, D. E. Octanol/ Water Partition Coefficients and Aqueous Solubilities of Organic Compounds; NBSIR 81-2406; U.S. Department of Commerce: Washington, DC, 1981. (68) Ahel, M.; Giger, W. Aqueous solubility of phenols and alkylphenol polyethoxylates. Chemosphere 1993, 26, 1461-1470. (69) Reed, H. W. B. Alkylphenols. In Kirk-Othmer Encyclopedia of Chemical Technology, Vol 2, 3rd ed.; Wiley & Sons: New York, 1978; pp 72-96. (70) Hill, D. J. T.; White, L. R. Enthalpies of solution of 1-hexanol and 1-heptanol in water. Aust. J. Chem. 1974, 27, 1905-1916. (71) Shinoda, K.; Yamanaka, T.; Kinoshita, K. Surface chemical properties in aqueous solution of nonionic surfactants: Octyl glycol ether, R-octyl glyceryl ether and octyl glycoside. J. Phys. Chem. 1959, 63, 648-650. (72) Urano, K.; Maeda, H.; Ogura, K.; Wada, H. Direct analytical method for aromatic compounds in water by steam carrier chromatography. Water Res. 1982, 16, 323-327. (73) Anderson, B. D.; Rytting, J. H.; Higuchi, T. Influence of selfassociation on the solubility of phenol in isooctane and cyclohexane. J. Am. Chem. Soc. 1979, 101, 5194-5197. (74) Casas, S. P.; Trejo, L. M.; Costas, M. Self-association of phenols in inert solvents. Apparent heat capacities of phenol, substituted phenols and aromatic alcohols in n-heptane. J. Chem. Soc., Faraday Trans. 1991, 87, 1733-1738.
Received for review April 14, 2005. Revised manuscript received September 4, 2005. Accepted September 7, 2005. ES050729D
