Environ. Sci. Technol. 2002, 36, 955-961
Evidence for Nonlinear Binding of PAHs to Dissolved Humic Acids YAEL LAOR* AND MENAHEM REBHUN Faculty of Civil Engineering, Environmental and Water Resources Engineering, Technion, Haifa 32000, Israel
Binding of pyrene, fluoranthene, and phenanthrene to dissolved humic acids (HA) was determined by the fluorescence quenching (FQ) and complexation-flocculation (CF) methods. Determinations by the CF method, using varying contaminant concentrations and a constant HA concentration, yielded nonlinear Freundlich-type isotherms (n ) 0.65-0.84). Experiments using both the CF and the FQ methods with varying HA concentrations and a constant contaminant concentration yielded curved “Stern-Volmer”type plots that also indicate nonlinear binding. These findings suggest that linear partitioning or site complexation in the presence of excess available sites cannot fully describe the interactions of hydrophobic compounds with dissolved humic material. Site-specific hydrophobic interactions at limited interior or external molecular surfaces may be considered.
Introduction Dissolved humic substances in subsurface and surface waters and in sedimentary porewaters have been shown to have an effect on fate and transport of hydrophobic contaminants (e.g., refs 1-6). “Dissolved” humic substances can be viewed as truly dissolved anionic macromolecules or as submicron colloids that are mobile in the subsurface environment, thus facilitating the transport of bound contaminants. On the other hand, “solid-phase” humic materials are integral constituents of soil and sediment solid matrixes and result in retardation of readily sorbed contaminants. Operationally, dissolved humic materials will not precipitate upon centrifugation and/ or will pass through a filter of a defined pore size (0.2-0.45 µm; 1, 7-9). The term binding has been used to describe the interactions of contaminants with dissolved humic materials, as contrasted with sorption, which has been used to describe contaminant interactions with solid-phase humic materials (6, 9). Although these two humic phases are operationally separated, the terminology does not necessarily represent two different mechanisms. Binding coefficients of PAHs to dissolved humic substances were found in several studies to be much higher than sorption coefficients to solid-phase humics (6, 9-11). It was suggested that in contrast to dissolved humic material, part of the sites in mineral-bound humic substances is not available, resulting in reduced contaminant sorption. It can be also that sorption kinetics are much slower than binding kinetics, such that true sorption equilibrium in solid-phase humic materials is not always achieved (9). On the other hand, Chiou et al. (12) found that dissolved humic material was less effective as compared to the bulk organic matter in concentrating organic solutes and suggested that it is mainly due to the high polarity of dissolved humic * Corresponding author present address: Institute of Soils, Water, and Environmental Sciences, The Volcani Center, Agricultural Research Organization, P.O. Box 6, Bet Dagan 50250, Israel. Phone: +972 3 9683698; fax: +972 3 9604017; e-mail:
[email protected]. 10.1021/es001996g CCC: $22.00 Published on Web 01/26/2002
2002 American Chemical Society
materials and to their smaller organic environment, making the partition interaction less favorable for the solute. Early works suggested a partitioning model to describe the interactions between hydrophobic organic contaminants and humic materials (e.g., ref 13). This uniform, concentration-independent distribution of a solute into a threedimensional matrix was suggested for both dissolved and solid-phase humic materials. Evidence for partitioning mechanism was taken to include linear isotherms, absence of competition in multisolute systems, and good relationships between sorption or binding coefficients and Kow (4, 10, 12). More recent studies, however, reported nonlinear sorption and bisolute competition for various hydrophobic contaminants with soil or sediment (solid-phase) organic matter, indicating that site-specific sorption mechanisms are important in addition to partitioning. A dual-mode sorption model including both a partitioning domain and a sitespecific domain has been found to successfully describe a number of sorption isotherms in solid-phase organic matter. This dual sorption nature has been attributed to (i) glassy and rubbery domains in organic matter (14-17); (ii) small amount of high surface area carbonaceous organic material (HSACM) admixed in a larger amount of solvent-like organic matter (18, 19); (iii) rigidity in organic matter due to intramolecular polar bonds or site-specific interactions at limited interior or external molecular surfaces (20, 21). None of the above models has been applied to dissolved organic matter, as such it is still widely believed that binding interactions to dissolved humic substances can be adequately described by a partitioning process or simply by site complexation in which a substantial excess of HA is present (22). Yet, there are a few studies in which nonlinear binding may be evident. For example, binding isotherms for benzo[e]pyrene and benzo[k]fluoranthene with dissolved Aldrich HA (23) and for benzo[k]fluoranthene with water-soluble soil organic matter (23, 24) deviated from linearity, although it was suspected by the authors to be an experimental artifact. Other evidence include nonlinear binding of pentachlorobenzene to dissolved HA (25) and lindane to dissolved fulvic acid (26). Although the mechanisms of binding to dissolved humic substances and sorption to solid-phase humic substances may not be entirely different, the method chosen for analysis is different by definition. For solid-phase humic substances, sorption analysis can be simply performed by separating the free (dissolved) contaminant from the sorbed contaminant by filtration or centrifugation. On the other hand, a different methodology is needed for dissolved humic substances in order to separate between two dissolved entities, the contaminant and the humic material. Various methods have been developed, among which are the dialysis (27), reversephase separation (28), solubility enhancement (12), and fluorescence quenching (FQ) (22). The various techniques may yield significantly different results (29, 30). The FQ method has become one of the most popular techniques because of its simplicity and elegance, but it is limited to fluorescent compounds and has been shown to overestimate binding coefficient values (29). This overestimation was also supported by the results obtained using the complexationflocculation (CF) method, recently developed by Laor and Rebhun (31). Using the CF method, this paper reports nonlinear binding isotherms obtained with three PAHs (Table 1) and dissolved HAs of three sources. In addition, slightly curved SternVolmer-type plots obtained by both the CF and the FQ methods were found to give another evidence for binding VOL. 36, NO. 5, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
955
TABLE 1. Chemical and Physical Properties of the PAHs Used in This Study pyrene fluoranthene phenanthrene
MW
Sw (mg/L)
log Kow
202.26 202.26 178.24
0.135 (32, 34) 0.265 (33) 1.290 (32, 33)
5.18 (32) 5.22 (35) 4.57 (32)
nonlinearity. Accumulation of nonlinear evidence, as first reported in this paper, suggests that partitioning or site complexation in the presence of excess available sites cannot fully describe the interactions of hydrophobic contaminants with dissolved humic materials.
Materials and Methods Reagents and Chemicals. Phenanthrene (Aldrich, Milwaukee, WI. 98% pure), pyrene (Aldrich, 99% pure), and fluoranthene (Fluka, 97% pure) were used as received. Concentrated stock solutions were prepared in methanol. Aqueous solutions were prepared by transferring 0.1-0.8 mL of PAH stock into 1 L of deionized distilled water (d2H2O) containing 0.5 mequiv/L NaHCO3. The solution was stirred for 1-2 h, and pH was adjusted to 6 with 0.1 N H2SO4. All PAH analyses were made by fluorescence spectrometer (Shimadzu RF-1501) with excitation/emission wavelengths (nm/nm) of 249/365, 333/ 390, and 281/464 for phenanthrene, pyrene, and fluoranthene, respectively. Soil HA reference 1R102H and peat HA reference 1R103H were obtained from the IHSS collection. Stock solutions (150180 mg/L OC (organic carbon)) were prepared by dissolving HA in d2H2O, slowly elevating the pH to 10-11 (0.1 N NaOH), and stirring for 30 min. The pH was then adjusted to 8 (0.1 N HCl); the solutions were stirred for 2 additional hours and then filtered through 0.2 polycarbonate filter (Poeretics, Livermore, CA). Aldrich HA (Aldrich, sodium salt) was prepared by first dissolving the HA in a small volume of 1 N NaOH, then diluting with d2H2O, adjusting the pH to 6, and stirring for 3 h. The solution was finally filtered through Whatman GF-A followed by 0.45 µm (Gelman GN-60). Total organic carbon (TOC) concentrations of the stock solutions were determined by TOC analyzer (Shimadzu 5000A). Aluminum sulfate (1.84 g/L Al2(SO4)3‚16H2O, AnalaR, England, 98% pure; besides 0.1% of inorganic impurities, the rest is most likely due to the water of crystallization) was used as a coagulant for binding experiments using the CF method. Binding Analyses. Binding was determined in batch experiments by both the FQ and the CF methods as detailed before (31). Briefly, 30 mL Corex tubes were filled with aqueous solution (pH 6) of PAH at concentrations between 5 and 75% of their water solubilities. Dissolved HA was added, and the tubes were shaken (70 strokes/min) in the dark at room temperature (23 ( 2 °C) for the desired time period. In FQ experiments, the fluorescence (and UV absorbance for “inner-filter effect” corrections) was then measured to account for the free and bound PAH fraction at each HA concentration. In CF experiments, alum and the desired amount of base were added to the equilibrated HA-PAH solution (after the binding stage); tubes were shaken 10 times by hand and then 20 min on a reciprocal shaker at 40 strokes/ min. After settling overnight, the supernatant was withdrawn from the top 2-3 cm of the tube for fluorescence measurements to account for the free PAHs as well as for UV measurements to account for any HA remaining in solution after flocculation (UV absorbance after flocculation was always very low, and inner-filter corrections were not needed). To directly compare the CF and FQ methods in one experiment, fluorescence and UV absorbance measurements were made before the addition of alum (FQ method), the 956
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 5, 2002
cuvette content was immediately returned back into the Corex tube, and alum was then added for CF analysis. Data obtained by the FQ method were fitted to Stern-Volmer equation (eq 4; below). Similarly, the results of the CF method were plotted as Ctotal/Cfree vs OC (eq 3), allowing a direct comparison between the two methods. All analyses were performed in duplicates or triplicates and compared with control tubes containing the specific PAH concentration to account for any loss during the course of the experiment (usually between 1 and 3% only; 31). Figures include average values of triplicates (except Figure 3, which shows average values of duplicates) and error bars for their standard deviations. Nonlinear regression analysis (Origin, version 6.0, Microcal, Northampton, MA) was used to determine parameters of the Freundlich equation.
Mathematical and Graphical Presentation of Binding Analyses To facilitate a comparison between the results obtained by the FQ and the CF methods, it is essential to visualize the mathematical and graphical relationships between the two ways of presentation. If linear behavior is assumed, then:
Cs ) KdocCfree
(1)
where Cs is the humic-bound contaminant concentration (w/w), Cfree is the free contaminant concentration (w/v), and Kdoc is the binding coefficient normalized to the dissolved humic OC content. Linear isotherm (Cs vs Cfree) is characterized by a straight line with an intercept of zero, and its slope represents the value of Kdoc. Equation 1 can be expressed as
Cbound [OC]
)
Ctotal - Cfree [OC]
) KdocCfree
(2)
where Ctotal and Cbound are the total and bound contaminant concentrations (w/v), respectively. Rearranging eq 2 yields
Ctotal ) 1 + Kdoc[OC] Cfree
(3)
Plotting Ctotal/Cfree against [OC] yields a curve with an intercept of 1 and a slope that represents the value of Kdoc. Equation 3 is identical to the form of Stern-Volmer equation used by Gauthier et al. (22) to determine Kdoc values by the FQ technique, provided that the fluorescence intensity is proportional to concentration of the free contaminant only:
F0 ) 1 + Kdoc[OC] F
(4)
where F0 and F are the fluorescence intensity of the contaminant in the absence (F0) and in the presence (F) of a quencher (OC). If nonlinear Freundlich-type binding interactions are expected, then
Cs ) KFCfreen
(5)
Similarly to the development in eqs 2 and 3, it can be expressed as
Ctotal - Cfree [OC]
) KFCfreen
(6)
and
Ctotal ) 1 + KF[OC]Cfreen-1 Cfree
(7)
FIGURE 1. Hypothetical linear and nonlinear binding relationships presented as Cs vs Cfree isotherms (a, c) and as Stern-Volmer type plots (b, d). Equation 7 is nonlinear and is not identical to the form of Stern-Volmer equation (eq 4). Therefore, the use of SternVolmer equation to calculate Kdoc values is valid only if binding is linear. Equation 4 was indeed derived by Gauthier et al. (22) simply from an equilibrium expression, which was eventually based on linear partitioning uptake or on site complexation in the presence of excess HA. Figure 1 illustrates these two ways of presentation for linear (panels a and b) and nonlinear (panels c and d) relationships. Each data point is numbered to illustrate the fact that the data point that is the closest to the origin on the Cs vs Cfree plot is the farthest from the origin on the equivalent Stern-Volmer plot. When binding is linear (Figure 1a,b), the data fit a straight line with an intercept of 0 for Cs vs Cfree plot or of 1 for Ctotal/Cfree vs OC plot. When binding is nonlinear (Figure 1c,d), the Cs vs Cfree plot is curved toward the x-axis, and the Ctotal/Cfree vs OC plot is curved toward the y-axis. Note that binding experiments that are performed by varying OC and maintaining a constant Ctotal can be represented either as Cs vs Cfree isotherms or as Stern-Volmer-type plots. However, for binding experiments that are performed by varying Ctotal and maintaining a constant OC, only Cs vs Cfree isotherms can be plotted since on Stern-Volmer-type plots such data will give a single data point in a linear case or will distributed vertically in a nonlinear case. In cases when Stern-Volmer-type plots are constructed for a narrow Ctotal/Cfree (or F0/F) range, curvature toward the y-axis (which indicates nonlinearity) may be masked. A “warning sign” for such curvature is that the intercept of the linear regression line falls below 1. This can be observed but has not been discussed in several studies in which the Ctotal/ Cfree range was up to 1.5-3 only (e.g., phenanthrene with several HAs, 9; pyrene with Suwannee HA, 29; pyrene with soil and peat HA, 31; and naphthalene with water-soluble soil organic matter, 36). By comparing Figure 1 panel a with panel b and panel c with panel d, it becomes clear that when Stern-Volmer plots are constructed only for such a narrow Ctotal/Cfree range, they represent only a small segment of the equivalent Cs vs Cfree isotherm. As such, it is critical to realize that linear or nonlinear behavior can hardly be deduced from most of these binding studies.
Results and Discussion Binding Analyses Using Varying Ctotal and a Constant OC Concentration (CF Method). A nonlinear binding isotherm
FIGURE 2. Binding isotherm of pyrene with soil HA (CF method) at contact time (before alum addition) of 15 min and 20-24 h. Total pyrene concentrations: 5-30% of its water solubility; OCHA, 5.9 mg/L. Apparent Kdoc values were calculated for each individual data point (inset).
FIGURE 3. Binding isotherms for pyrene with peat HA, fluoranthene with Aldrich HA, and phenanthrene with Aldrich HA (CF method). Contact time: 20-24 h (before alum addition). Total concentrations (in respect to compound water solubility): 10-45% (pyrene), 1040% (fluoranthene), and 20-75%, (phenanthrene); OCHA, 17.5 mg/L. (n ) 0.65 ( 0.02) was obtained for pyrene with dissolved soil HA at contact times of 15 min and 20-24 h (Figure 2). This contact time represents the “complexation” stage before alum is added (31). Nonlinearity can be clearly seen in the inset of Kdoc values calculated for each individual data point plotted against Cfree values. The strong decrease in apparent Kdoc values with increasing solute concentrations proves nonlinear binding behavior. Nonlinear binding isotherms were also observed for pyrene with dissolved peat HA (n ) 0.68 ( 0.005) and for phenanthrene (n ) 0.73 ( 0.042) and fluoranthene (n ) 0.69 ( 0.039) with dissolved Aldrich HA (Figure 3). All linear and nonlinear Freundlich coefficients values are summarized in Table 2. Binding Analyses Using Varying OC Concentrations and a Constant Ctotal (FQ and CF Methods). Stern-Volmer-type plots obtained for pyrene with dissolved soil HA in the twostage experiment (FQ followed by CF analysis) are presented in Figure 4. Higher binding was obtained by the FQ analysis as previously reported (31). Danielsen et al. (29) also found VOL. 36, NO. 5, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
957
TABLE 2. Binding Coefficients (Linear and Nonlinear) Obtained in This Study and Linear Binding Coefficients from Literature (( Standard Error)a method Fig
Kdoc (linear)b
KF (non linear)c
CFd CFe CF f FQ CF FQ
2 5 4 4 3 6
2.03 × 105 ((9.41 × 103) 3.61 × 105 ((19.76 × 103) 2.16 × 105 ((10.42 × 103) 4.90 × 105 ((26.13 × 103) 1.70 × 105 ((16.71× 103) 1.77 × 105 ((4.92 × 103)
Pyrene 6.20 × 102 ((30.50) 5.89 × 102 ((50.13) 6.49 × 102 ((2.39) 8.84 × 102 ((137.67) 4.23 × 102 ((46.74) 2.70 × 102 ((29.82)
0.65 ((0.020) 0.70 ((0.033) 0.68 ((0.001) 0.67 ((0.059) 0.68 ((0.005) 0.84 ((0.035)
Aldrich HA
CF FQ
3 6
7.51 × 104 ((7.10 × 103) 1.15 × 105 ((3.32 × 103)
Fluoranthene 2.58 × 102 ((35.99) 2.34 × 102 ((35.07)
Aldrich HA
CF FQ
3 6
1.22 × 104 ((1.07 × 103) 4.10 × 104 ((3.32 × 103)
IHSS soil HA
IHSS peat HA
Phenanthrene 75.68 ((20.21) na
Kdoc literature (linear)b
n (nonlinear
}
2.47-3.43 × 105 (ref 41, FQ) 0.88-3.2 × 105 (ref 47, FQ) SH
}
not found
0.69 ((0.039) 0.79 ((0.040)
}
1.84 × 105 (ref 6, DL) 2.10 × 105 (ref 6, FQ)
0.73 ((0.042) na
}
4.57 × 104 (ref 48, FQ)
a FQ, fluorescence quenching; CF, complexation-flocculation; DL, dialysis; SH, soil HA (other than IHSS); na, not applicable. b Units in mL/g of OC. c Units in [(µg of PAH/g of OC)/(µg of PAH/L)n]. d Varying pyrene concentration. e Varying pyrene concentration and varying HA concentration. f Varying HA concentration.
FIGURE 4. Stern-Volmer-type plots for pyrene with soil HA obtained by the FQ and CF methods. Total pyrene concentration: 25% of its water solubility. Contact time: 20-24 h (before alum addition). Although linear relationships are observed (solid linear fit lines), the plots are curved toward the y-axis as shown by the dotted lines. Apparent Kdoc values were calculated for each individual data point (inset). that binding coefficients for pyrene with dissolved HA were overestimated by the FQ method compare with the solubility enhancement approach. They postulated that it was resulted from the fact that pyrene was quenched not only statically by associating with the HA (which is the theoretical basis of the FQ method) but also dynamically by molecular oxygen. In the present study, the differences between the Kdoc values obtained by the FQ and CF methods are by factors of 1.042.41 (pyrene), 1.53 (fluoranthene), and 3.36 (phenanthrene). In our former study (31), the differences were by factors of 1.13-1.35 (pyrene), 1.24 (fluoranthene), 2.79-3.54 (phenanthrene), and 2.04 (anthracene). These Kdoc overestimations are comparable with the results of Danielsen et al. (29), who reported differences between the FQ and the solubility enhancement methods for pyrene with HA by a factor of 1.58, and the results of Peuravuori (37), who reported an average factor of 2.29 for the interactions of pyrene with HA of various sources. Higher factors were reported by these authors for pyrene interactions with fulvic acids (29, 37). More interesting in the context of this study is the fact that although significantly linear (r2 ∼0.99) both plots in Figure 4 are slightly curved toward the y-axis. This phe958
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 5, 2002
nomenon can actually be seen (but is not discussed) in our former study when the CF method was first presented (31). Puchalsky et al. (38) also recognized such curvature in SternVolmer plots obtained by FQ analyses and related it to the pseudomicelle structure of HA in which the fluorescent contaminant is restrained but not rigidly bound. They suggested that in such a situation the quenching behavior of HA does not conform to a simple Stern-Volmer theory and curved plots are obtained. This interpretation needs further proof as discussed in the section dealing with possible experimental artifacts. The present study shows that curved plots can be obtained simply because nonlinear binding relationships exist (Figure 1c,d). It still could be that such curvature can be the result of both deviation from SternVolmer theory (38) and nonlinear binding effects. The fact that such curvature is also obtained when the analytical method is based on a completely different principle, i.e., complexation- flocculation process, gives further support for a nonlinear behavior. It is important to realize that since these apparent linear Kdoc values depend on the range of solute concentrations, they cannot be directly compared with Kow values. It is clearly seen in Figure 4 that when binding coefficients are simply derived from the “linear” slope of curved Stern-Volmer plots, then the derived Kdoc will depend on the range of values measured for Ctotal/Cfree (or F0/F), with greater ranges yielding higher Kdoc values. Converting Stern-Volmer-Type Plots into Cs vs Cfree Binding Isotherms. As shown by eqs 1-7, data that are represented by Stern-Volmer-type plots can also be represented as Cs vs Cfree isotherms. The data represented as a Stern-Volmer plot for pyrene with soil HA (Figure 4) were converted into Cs vs Cfree isotherms, and Kdoc values were then determined for each individual data point (inset). The converted data of the CF method (lower curve in Figure 4) are shown in Figure 5 together with data that are presented in Figure 2 for the experiment where Ctotal was varied (20-24 h contact time). Both sets of data fall essentially on one significantly nonlinear Freundlich isotherm (n ) 0.70 (0.033). The insets are shown to illustrate the decrease in apparent Kdoc over the experimental range of Cfree values. Stern-Volmer plots obtained by FQ analysis for pyrene with peat HA and for fluoranthene and phenanthrene with Aldrich HA (Figure 6a) were also curved toward the y-axis (intercept
