Environ. Sci. Technol. 1997, 31, 1703-1710
A Distributed Reactivity Model for Sorption by Soils and Sediments. 9. General Isotherm Nonlinearity and Applicability of the Dual Reactive Domain Model WEILIN HUANG, THOMAS M. YOUNG,† MARK A. SCHLAUTMAN,‡ HONG YU,§ AND WALTER J. WEBER, JR.* Environmental and Water Resources Engineering, Department of Civil and Environmental Engineering, The University of Michigan, Ann Arbor, Michigan 48109-2125
Sorption isotherms for a hydrophobic solute probe, phenanthrene, were measured experimentally for 27 different soils and sediments. The linear and Freundlich isotherm models and the Dual Reactive Domain Model (DRDM) were used to fit the resulting data. The results reveal for all soils and sediments studied that (i) the Freundlich model and the DRDM fit the data well, whereas a linear model fails to do so; (ii) values of the organic carbon-normalized distribution coefficient, KOC, calculated from individual isotherm points for a specific sorbent-solute system vary significantly with the aqueous-phase solute concentration, Ce; and (iii) all commonly used correlations of KOC with octanolwater partitioning coefficients and solute solubility limits significantly underestimate KOC for Ce values smaller than approximately one-tenth of aqueous-phase solute solubility, CS. The sorption behaviors of all of the soils and sediments studied are thus inconsistent with the simple concept of linear phase partitioning. The general applicability of the DRDM, a polymer-based limiting case form of the Distributed Reactivity Model, for all systems studied supports mechanistic arguments based on polymer sorption theory.
Introduction Previous studies have revealed that fate and transport models for hydrophobic organic contaminants (HOCs) in environmental systems involving soils, sediments, and/or aquifer materials are highly sensitive to sorption isotherm nonlinearity (1-5). That sorption on natural solids comprising diagenetically altered organic matter (e.g., kerogens) yields nonlinear sorption isotherms has been broadly confirmed (610). The case for sorption on natural solids comprising geologically younger humic materials has not been as clear however. Observations of nonlinear behavior in some instances (1, 2, 6-8, 11-16) and linear behavior in others (17-24) have been variously invoked as evidence to support different mechanistic interpretations and arguments. * Corresponding author e-mail:
[email protected]; phone: 313-763-1464; fax: 313-763-2275. † Present address: Department of Civil and Environmental Engineering, University of California, Davis, CA 95616. ‡ Present address: Environmental, Ocean, and Water Resources Division, Department of Civil Engineering, Texas A&M University, College Station, TX 77843-3136. § Present address: Department of Microbiology and Immunology, School of Public Health and Nursing, Guiyang, Guizhou, People’s Republic of China.
S0013-936X(96)00677-3 CCC: $14.00
1997 American Chemical Society
Several investigators have advanced thermodynamic justifications for a simple phase partitioning model (19, 2528). Karickhoff (19, 25), for example, used a fugacity approach to reinforce a theoretical argument for linear partitioning by assuming that activity coefficients in the aqueous and sorbed phases were independent of solute concentration, but noting in his discussions that nonlinear partitioning could result if activity coefficients vary with solute concentration. Spurlock and Biggar (27, 28) expanded the concept of a generalized partitioning model predicated on nonconstant sorbed-phase activity coefficients, arriving at a relationship mathematically equivalent to a modified Freundlich equation, and invoked the Flory-Huggins theory to predict the sorbed-phase activity coefficients. As noted by Young and Weber (8), however, such nonlinear partitioning models do not account for the ranges and magnitudes of sorption nonlinearity and capacity often observed for different soils and sediments with respect to less polar and nonpolar HOCs. Recent studies have revealed that sorption of slightly polar (e.g., chlorinated solvents) and nonpolar (e.g., polynuclear aromatic hydrocarbons) HOCs by soils and sediments is highly dependent on the chemical and structural properties of associated soil organic matter (SOM) (6-8, 29-31). The more chemically reduced and physically condensed nature of kerogen-type SOMs derived from sedimentary parent rocks, for example, are found to have much higher sorption capacities and more nonlinear isotherms than geologically younger humic-type SOMs (6-8, 10, 31). Previous research has shown that SOM may comprise two principal types of heterogeneous organic domains, a “hard” carbon domain and a “soft” carbon domain (6-8, 11, 12, 32-35). The soft carbon domain can be envisioned as a highly amorphous and swollen SOM, analogous to a rubbery polymer (33). The hard carbon domain can be envisioned as a condensed and relatively rigid organic matrix, analogous to a glassy polymer (33). If these analogies are accurate, it is expected from the known respective sorption behaviors of the rubbery and glassy states of polymers that the two different types of SOM domains will exhibit different sorption and desorption rates and equilibria for less polar and nonpolar HOCs. The condensed or glassy domain, for example, would be expected to exhibit nonlinear sorption behavior, slower rates of sorption, solutesolute competition, and possible sorption-desorption hysteresis. Conversely, the highly amorphous or rubbery domain should exhibit linear sorption behavior, faster rates of sorption, no solute-solute competition, and no sorption-desorption hysteresis. While the concept of two principal types of SOM domains appears quite reasonable as a basis for interpreting sorption and desorption behaviors commonly observed for natural sorbents, further experimental evidence is required to confirm the concept and to establish a mechanistic model having predictive capability. It is our purpose in this study to compile the experimental evidence required to support or refute convincingly the dual SOM domain concept and the associated polymer-based Dual Reactive Domain Model (DRDM) advanced by our research group in the preceding paper in this series (33). To this end we have (i) investigated the degree of isotherm nonlinearity of phenanthrene, a nonpolar HOC probe on a large number and broad range of different soils and sediments; (ii) examined the suitability of several models for describing the observed equilibrium sorption behaviors; and (iii) quantified the variability of the commonly invoked organic carbon-normalized distribution coefficient, KOC, for the soils and sediments tested. A systematic experimental approach involving 27 wellcharacterized natural solids comprising different quantities and types of SOM, inorganic components, and different
VOL. 31, NO. 6, 1997 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1703
TABLE 1. Sorbent Characteristics sample
BET-N2 surface area (m2/g)
TOC (wt %)
sanda (wt %)
silta (wt %)
Clay a (wt %)
CECa cmol(+)/kg
Houghton Muck (soil, MI) Webster (soil, IW) Chelsea I (topsoil, MI) Chelsea II (topsoil, MI) EPA-B2 (stream sediment, GA) EPA-4 (river sediment, ND) EPA-5 (river sediment, ND) EPA-6 (river sediment, SD) EPA-8 (river sediment, IA) EPA-9 (loess, IA) EPA-13 EPA-14 (soil, WV) EPA-15 (river sediment, IN) EPA-18 (river sediment, KY) EPA-20 (soil, IL) EPA-21 (river sediment, IL) EPA-22 (river sediment, IL) EPA-23 (lake sediment, IL) EPA-26 (river sediment, IL) Ann Arbor II (glacial till, MI) Wagner II (aquifer sand, MI) EPA-14-BEe EPA-15-BE EPA-20-BE EPA-22-BE EPA-23-BE Chelsea II-BE
0.97 9.46 3.92 3.92 8.20 21.80 15.70 59.70 5.34 19.80 12.48 41.60 15.24 ND 10.23 6.96 3.25 33.20 16.15 3.58 1.43 ND ND ND ND ND ND
46.24 2.97 5.60 5.45 ( 0.10c 1.07 ( 0.03 2.28 ( 0.05 2.07 ( 0.14 0.79 ( 0.02 0.158 ( 0.011 0.118 ( 0.004 4.50 ( 0.10 0.446 ( 0.031 1.24 ( 0.09 0.543 ( 0.030 1.18 ( 0.03 2.36 ( 0.16 1.65 ( 0.06 2.57 ( 0.08 1.50 ( 0.01 0.561 ( 0.020 0.151 ( 0.013 0.241 ( 0.003 1.11 ( 0.08 0.600 ( 0.002 1.29 ( 0.14 1.52 ( 0.06 2.44 ( 0.09
NDb ND ND ND 67.5 3.0 33.6 0.2 82.4 7.1 ND 2.1 15.6 34.6 0 50.2 26.1 17.3 1.6 ND 100d ND ND ND ND ND ND
ND ND ND ND 13.9 41.8 35.4 31.2 10.7 75.6 ND 34.4 48.7 25.8 71.4 42.7 52.7 13.6 55.4 ND 0 ND ND ND ND ND ND
ND ND ND ND 18.6 55.2 31.0 68.6 6.8 17.4 ND 63.6 35.7 39.5 28.6 7.1 21.2 69.1 42.9 ND 0 ND ND ND ND ND ND
ND ND ND ND 3.72 23.72 19.0 33.01 3.72 12.40 ND 18.86 11.30 15.43 8.50 8.33 8.53 31.15 20.86 ND ND ND ND ND ND ND ND
a
Based on refs 20 and 36.
b
Not determined. c (1 SD.
d
From this study. e BE designates base-extracted sample.
particle or aggregate size distributions was employed. Sorption equilibria were examined over a broad spectrum of residual aqueous-phase solute concentrations. Relative isotherm linearities for the sorbent-solute systems studied were evaluated on the basis of goodness-of-fit of each of three sorption isotherm models (linear, Freundlich, and DRDM), the value of the exponent term (n) in the Freundlich model, and the dependence of the linear distribution model coefficient (KD,L) of individual isotherm points on residual aqueous-phase solute concentration. The goodness-of-fit of the DRDM isotherm to the experimental data was used to assess the suitability of the linear/nonlinear composite SOM concept for interpretation of sorption mechanisms.
Experimental Section Sorbents. Twenty-seven soil and sediment samples representing a spectrum of heterogeneous natural solids were used in this study. Twenty-one of these were examined in their natural state, including a muck soil from Michigan (Houghton), three Michigan top soils (Chelsea I, Chelsea II, Ann Arbor II), a silty clay loam topsoil from Iowa (Webster), a sandy subsurface material containing shale particles from Michigan (Wagner II), and 15 reference EPA soil and sediment samples obtained from W. L. Banwart of the Department of Agronomy at the University of Illinois at Champaign-Urbana. These natural solids were selected because of their well-characterized properties and the fact that they have been employed in previous studies of organic contaminant sorption under both equilibrium and rate-limited conditions, either by our research group (6-8, 11) and/or by others (9, 19-21, 23, 25, 36-38). In addition, base-extracted samples of six of the 21 natural solids were examined; five of the EPA samples (EPA14, -15, -20, -22, -23) were extracted four times, and the Chelsea II was extracted 16 times using 0.01 N NaOH at a soil-solution ratio of 20:250 (w:w) under a nitrogen atmosphere. Each sorbent was characterized in terms of its N2-BET specific surface area and its total organic carbon (TOC)
1704
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 31, NO. 6, 1997
content. Surface areas were evaluated using nine-point nitrogen gas adsorption data collected at liquid nitrogen temperature (77.35 K) for all original samples (ASAP-2010, Micromeritics). TOC contents were analyzed using a hightemperature combustion method (CHN-1000 analyzer, Leco, Inc.) on the original and base-extracted samples after pretreating with 0.5 M hydrochloric acid. The results, based on triplicate analyses, are listed in Table 1, along with selected sorbent characteristics. Solute. Phenanthrene, a three-ring polynuclear aromatic hydrocarbon (PAH), was selected as a solute probe because of its common association with aquatic and subsurface contamination resulting from past petroleum and coal production practices and because of its extensive prior use in the DRM series of investigations (8, 11, 33, 39). Spectrophotometric grade (>98%) phenanthrene was obtained from Aldrich Chemical Co., Inc. A background solution comprising 0.005 M CaCl2 and 100 mg/L NaHCO3 was employed. Sodium azide (NaN3) was added at a level of 100 mg/L to control biological activity. Methods used for preparation of stock and initial aqueous solutions have been described previously (11). Sorption Experiments. Established procedures (8, 11, 39) employing a bottle-point, fixed-sorbent dosage technique were utilized for conducting all equilibrium sorption experiments in completely-mixed batch reactors (CMBRs) operated at 25 ( 0.3 °C. Three types of glass CMBRs were employed: (1) 35-mL centrifuge bottles for the Chelsea I, II, base-extracted Chelsea II-BE, and Webster samples; (2) 125-mL bottles for Houghton muck; and (3) 20-mL flame-sealed glass ampules for all other samples. The 35-mL reactors were capped with Teflon-lined septa, and the 125-mL reactors were capped with silver foil and Teflon-lined septa, with the foil contacting the solution phase. Based on preliminary sorption rate and equilibrium studies and depending on the particular sorption characteristics of each sorbent, an equilibration time in the range from 14 to 28 days and a constant soil-water weight ratio in the range from 1:600 to 1:4 were chosen to achieve 30-70% reductions in initial solute concentration for each
sorbent-solution system. Initial aqueous-phase solute concentrations were selected to yield a set of isotherm data for each sorbent that was distributed evenly on a log-log scale plot and spanned approximately 2 orders of magnitude in residual aqueous-phase solute concentration. Reactors were mixed completely by tumbling top to bottom at 12 rpm. The suspended solids were separated from the aqueous phase after equilibration by centrifugation at 2000 rpm for 20 min for the centrifuge bottle reactors and by sedimentation for a period of 2-3 days for the other types of reactors. A preliminary study showed no difference between the two techniques. After separation, an aliquot of supernatant was withdrawn and mixed with a predetermined amount of methanol in a 5-mL glass vial that was then capped with a Teflon-lined septum. The mixture was analyzed for phenanthrene using reverse-phase HPLC (ODS, 5 µm, 2.1 × 250 mm column on a Hewlett-Packard Model 1090) with a diode array detector for concentrations ranging from 50 to 1000 µg/L and a fluorescence detector for concentrations from approximately 0.5 to 50 µg/L. The aqueous-phase solute concentration was calculated by applying a dilution factor computed from the density data of the methanol-water mixture to the actual analytical measurement (11). The solid-phase solute concentration was determined from a solute mass balance between the two phases. Solute loss to reactor components (∼3-7%) was taken into account for the 35-mL bottle reactors during mass balance calculations using results from simultaneously operated sorbent-free control reactors. Solute losses for the 20-mL glass ampules and the 125-mL bottles were found to be insignificant ( 0.971); and (iii) the linear model is not appropriate for describing any of the experimental sorption data obtained. Figures 1-3 demonstrate the fitting of the isotherm data for the Houghton muck soil by the three different sorption models. It is clear from these figures that the Freundlich model (eq 2) fits the equilibrium sorption data much better than the linear model (eq 1) over the entire concentration range examined (Figure 1) as well as for the narrower concentration range of Ce < 100 µg/L (Figure 2). The results shown in Figure 2 directly contradict the widely-held belief that linear isotherms prevail at low solute concentrations; i.e.,
