Environ. Sci. Technol. 1994, 28, 059-067
Adsorption/Desorption Hysteresis in Organic Pollutant and SoiVSedirnent Interaction Amy T. Kan,. Gongmln Fu, and Mason B. Tomson
Department of Envlronmental Science and Engineering, Rice University, Houston, Texas 77251 Adsorption and desorption of pollutants to soil and sediment materials are major fate mechanisms. The hypothesis that adsorption and desorption are reversible processes has been tested. The organic pollutants naphthalene, phenanthrene, and p-dichlorobenzene have been studied in the laboratory using batch reactors at room temperature from a few hours to over 2 months. The adsorption experiments were at equilibrium within 1-4 days and could be modeled using simple linear isotherms with K, values consistent with published KO,and KO, relationships. Desorption experiments were conducted with the contaminated sediments by successive dilutions. Desorption experiments varied from 1day to 5 months, and observed desorption rates were from 1to 3 orders of magnitude smaller than previously measured or predicted. If equilibrium were obtained during the desorption, typically over 82-99% of the adsorbed pollutant would have been desorbed, but generally only 30-50% of the adsorbed pollutant could be desorbed. These desorption results could not be explained by commonly invoked kinetic models or artifacts of the procedure. The possibility and consequencesof such adsorption and desorption behavior being the result of either hysteresis or irreversible adsorption is discussed.
Introduction Adsorption/desorption is often an important mechanism in the fate and transport of neutral hydrocarbon pollutants in the environment. In modeling pollutant transport, the adsorption/desorption process is often simplified by assuming what might be called ideal conditions of instantaneous equilibrium, isotherm linearity and desorption reversibility (1).However,many field and laboratory data deviate from that predicted by this simple model approach; as a consequence, the assumption of ideal behavior has been challenged by numerous researchers (1-7) and is the primary focus of the present paper. Adsorptionldesorption behavior, which does not conform to such a simple idea, has been attributed to several different factors, including the following: (a) varying adsorption energies, leading to isotherm nonlinearity (i.e., a Freundlich isotherm, see refs 1 and 5-8); (b) failure to attain equilibrium in either the adsorption or the desorption directions due to slow kinetics in either step (3, 4, 8-10); (c) chemisorption of the pollutants to various components of the soil matrix, causing apparently irreversible adsorption (1); (d) either biotic or abiotic degradation of the pollutant being studied, again causing an apparently irreversible adsorption (11,l.Z);(e) adsorption/ desorption hysteresis (1, 13, 14); and (f) experimental procedures, such as centrifugation versus dilution (15). Only those ideas directly related to this research paper will be reviewed herein, but refs 1-4, 8, and 16-18 are recommended for further reading. Adsorption nonlinearity has been studied extensively (I, 5,11,19),and the effect of this nonlinearity on pollutant 0013-936X/94/0928-0859$04.50/0
0 1994 American Chemical Soclety
transport has been investigated by several researchers (8, 9, 20). Adsorption nonlinearity is commonly fitted to a Freundlich isotherm, with an exponent less than unity, although numerous other nonlinear isotherms are feasible (21, 22). If the true adsorption mechanism is such that it is correctly described by a Freundlich isotherm with an exponent less than one, the consequencesconcerningbatch adsorption kinetics, transport, and fate can be substantial. For example, the leading edge of a pollutant breakthrough in groundwater will be substantially retarded versus that which would be predicted using a linear isotherm (8, 9). Similarly, the trailing edge of the flush-out curve would be much longer than that predicted using a linear isotherm. There would be a corresponding concentration-dependent effect upon several types of kinetics equations, e,g., those which depend upon the partition coefficient (K,)and pore diffusion models. In practice, along with mechanistic considerations, the use of a Freundlich isotherm instead of a simple linear isotherm often causes the mathematics associated with the kinetics or with the transport modeling to be considerably more difficult. Also, there is no generally accepted procedure to estimate either the exponent or the partition coefficient for Freundlich isotherms as there is for linear isotherm partition constants. Therefore, there is a commensurate reluctance to use such a mathematical description, unless its use can be justified strongly (23). With hydrocarbon pollutants and sediments, it is often observed in the laboratory that the desorption isotherm is not the same as the adsorption isotherm, and this difference is the basis for suggesting the existence of hysteresis. This type of hysteresis has been reported for several classes of compounds,including polycyclicaromatic hydrocarbons, chlorinated benzenes, pesticides, phenols, surfactants, halogenated aliphatic hydrocarbons, and PCBs (6,7,13,17,24,25). Several authors have suggested that such reports of hysteresis are in fact generally a consequence of slow adsorption or desorption kinetics (1, 3, 4, 10). Karickhoff and Morris (26) used a twocompartment model, with equal adsorption intensity in each compartment, to account for the observed slow approach to equilibrium from either direction (portions of this model will be used later in the Results and Discussion section of this paper). Several empirical correlations have been used to relate rate constants to the partition coefficient (26-28). Models of radial pore diffusion have also been used to account for the slow kinetics. These radial diffusion models typically differ in the way they treat the effective diffusion coefficient, the porosity, and the tortuosity of the sediment particles (24,8-10,291. Adsorption and desorption time of minutes to years are predicted using these various kinetic approaches. To date, too few sedimentlhydrocarbonsystems have been studied experimentally over a sufficient range of variables, such as OC content and type, soil types and states of aggregation, pH and redox potential, temperature, and ionic strength and composition, to mention a few. In addition, changes in these parameters might affect the Environ. Sci. Technol., Vol. 28, No. 5, 1994 859
observed adsorption and desorption kinetics (26). The concept and existence of “true” hysteresis in adsorption and desorption have been discussed (1,14,16, 30,311. Flanagan et al. (31) state that “The existence of hysteresis ... in contrast to other irreversible processes, the extent of irreversibility is repeatable through many cycles of hysteresis”. This type of “true” hysteresis has been documented for several gadsolid adsorption systems (30,311for surfactant adsorption to solids (13,25,30,31). Adamson (16) suggests that there are three major types of hysteresis loop shapes, two types of closed-loop hysteresis in which the desorption is 100% complete, and one type in which a fraction of the adsorbate is irreversibly bound to the adsorbent (even at zero solution-phase concentration). The closed-looptype of hysteresis is often modeled based upon different shapes of capillary tubes in the adsorbent, such as the classical ink-bottleshaped pores, etc. (16). Most of these discussions are based upon the adsorption of gases to various solids, but the same arguments can be applied to the adsorption of neutral hydrocarbons from aqueous solution onto porous sediment particles. Adamson (16) suggests that the open-loop type of hysteresis is probably due to a mechanical or structural rearrangement of the adsorbent, i.e., the solid from which desorption takes place is different from that during adsorption (see also ref 1). The latter type of open-loop shaped hysteresis might more appropriately be called simply irreversible adsorption. Seri-Levy and Avnir (32) suggested that open-loop hysteresis could also be caused by heterogeneous surface geometry effects. From these references, three types of anomalous adsorption/desorption behavior can be classified: (1)apparent hysteresis, which is the result of some experimental artifact; (2) true hysteresis, which is time invariant and repeatable; and (3) irreversible adsorption, which can have different explanations depending upon the specific system, but is generally associated with some rather permanent change in the adsorbent/adsorbate system. Methylene blue adsorption is commonly used to determine the surface area of solids in situ. Surprisingly, its desorption from nonporous anatase (TiOz) was found to be irreversible (13),and the authors emphasized the importance of doing desorption studies when characterizing adsorbate/adsorbent interactions. In studying hydrocarbon adsorption to sediments, Karickhoff and Morris (26) used a twocompartment model (one fast and one slow compartment) to explain the observed slow approach to adsorption equilibrium, but emphasized that no one knows either where the soil organic carbon is located or where the adsorbed hydrocarbon molecules are located in the soil matrix and that all such models assume that there are no changes in the soil matrix during adsorption and desorption. These concerns could apply to pore diffusion models that are also used to account for slow adsorption/ desorption. Soil scientists have observed that the longer a pesticide is applied to a soil, the greater is the fraction of pesticide that is irreversibly bound (1, 7, 33, 34). Recently, 1,2dibromoethane (EDB), a soil fumigant, was found in agricultural topsoil 19 years after its last known application, despite its high volatility and degradability (34). Similarly,a fraction of sedimenbbound polycyclic aromatic hydrocarbons (PAHs) from Tamar Estuary, U.K., was found to be inert to leaching, microbial degradation, or photodegradation (35). Using presently accepted mechanisms of adsorption and diffusion, it is often difficult to 860
Envlron. Scl. Technol., Vol. 28, No. 5, 1994
account for the observed persistence of pollutants. In many of these cases, various mechanisms that could lead either to apparent hysteresis, true hysteresis, or irreversibility may be responsible for the observed slow release of pollutants. It is expected that the opportunity for irreversibility to occur will increase with exposure time, due to the slow weathering of the soil and due to the environmental changesfrom infiltration, etc. that can cause the soil matrix to change. On the other hand, Brusseau and Rao ( I ) state that “HOCs [hydrophobic organic chemicals] are expected to exhibit singularity [i.e., the absence of true hysteresis] under most conditions.” Furthermore, they suggest that if true hysteresis occurs, it will likely be insignificant or can be incorporated into the transport model. In summary, there is still uncertainty about how best to model pollutant adsorption/desorption, and numerous additional studies will be needed to resolve the issues. The overall objective of this research has been to determine factors that influence the release of pollutants from sediments. Batch adsorption followed by successive desorption experiments were used. The experiments were designed to minimize many of the experimental artifacts, which are known to cause apparent hysteresis. The priority pollutants naphthalene and phenanthrene with well-known physical chemical properties were chosen as model compounds. Experimental Section
Sorbent. The sorbent used was sediment obtained from Johnson Ranch, Lula, OK, where the ranch is located near the margin of the flood plain of a small river. The sediment consists of 92% sand, 6% silt, and 2% clay (36). It was air-dried, sieved through cheesecloth to remove vegetative matters and pebbles, and stored in the refrigerator. Since this is a surface sediment, trace amounts of plant debris (e.g., roots) still remained in the sample after sieving. No obvious adsorption interference from the plant debris was observed. The organic carbon content was determined with the LECO combustion method corrected for inorganic carbonates (GalbraithLaboratories, Inc., Knoxville,TN). According to Clark ( 3 3 ,the sediment granule sizes ranged from 0.2 mm to less than 0.063 mm (nominal dimension of the sieve opening). About 75% by weight of the sediments are fine sand grains with a diameter between 0.11 to 0.5 mm and a weighted average diameter of 0.23 mm. It contains 0.27% soil organic matter, evenly distributed among the larger sand grains (0.17-0.29% OC for 0.1-2.0 mm fractions), except for the silt and clay fraction. The clay fraction has a little higher organic carbon content (0.39-0.75 % OC for < 0.088-mm fractions). The BET surface area is 1.24 m2/g, determined with nitrogen as an adsorbing gas (Micromeritics, Inc). Sorbates and Chemicals. Radiolabeled P*CInaphthalene and [l%]phenanthrene (SigmaChemicalCo.) with a specific activity of 6.8 and 8.3 mCi/mmol, respectively, were used as the sorbate. The purity of these radiolabeled compounds was greater than 98%, as determined by GC/ MSD. Upon receipt of the radiolabeled naphthalene, a stock solution of 10 mCi/L (or 188 mg/L) was prepared in methanol. Chemicals used in this research were reagent-grade or better. Most naphthalene experiments used electrolyte solutions of various ionic strengths prepared with sodium chloride and deionized water. The solution also contained 0.0125 M NazB40710H20 (Mallinckrodt)and was adjusted
Table 1. Summary of 23 Naphthalene Adsorption Experiments experiment no.a adsorption time (day) 1.s. (M) la 2a 3a 4a 5a 6a 7a 8a 9a 10a lla 12a 13a 14a 15a 16a 17a 18a 19a 20a 21a 22a 23a I1
1 7 30 1 1 1 1 1 1 1 7 7 7 7 7 4 1 1 1 1 1 1 1
1
0.1 0.1 0.1 0.05 1.0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.005
Cinitialb
(rg/mL)
1.137 1.043 1.067 1.112 1.096 0.832 0.409 0.202 0.102 0.041 0.838 0.417 0.205 0.103 0.041 0.243 0.554 0.522 0.225 0.225 0.225 0.225 0.225 1.318
Cadsb
(rg/mL)
0.955 0.885 0.898 0.968 0.941 0.704 0.342 0.169 0.084 0.034 0.695 0.351 0.166 0.082 0.032 0.205 0.478 0.445 0.194 0.188 0.191 0.187 0.189 1.138
(?adsorbedc
(pg/g)
2.275 1.975 2.113 1.800 1.938 1.600 0.838 0.413 0.225 0.088 1.788 0.825 0.488 0.263 0.113 0.475 0.950 0.962 0.388 0.462 0.425 0.475 0.450 2.165
qextractsdd
(rg/g) mass balancee (%)
1.779 1.853 2.045 1.793 2.114
98.5 99.5 100.9 100.9 100.3
NAf NA NA NA NA
NA NA NA NA NA
1.577 0.918 0.477 0.256 0.099 0.444 1.010 1.041
99.9 100.0 101.3 99.3 87.4 99.8 99.4 99.6
NA NA NA NA NA NA
NA NA NA NA NA NA
Experiments were conducted at pH 8.0 in electrolyte solutions at various ionic strength adjusted with sodium chloride except in experiment 16a,where the solution medium also contained 0.01 M of CaCl~2Hz0, and in experiment 11,where the solution medium was 0.001M CaCl~2Hz0, 0.001 M NaHC03, and 0.0004 M MgCl. C h i t i d is the solution-phase naphthalene concentration before the adsorption experiment, and Cads is the solution-phase naphthalene concentration at the end of adsorption experiments. Qadsorbed is calculated by eq 1. q e x b a dis the amount of naphthalene extractable/g of sediment by methanol at the end of the adsorption experiment. e Mass balace = q e ~ r a e t e d / q a f~ NA r ~ . = not analyzed.
to pH 8.0 with HC1. Formaldehyde (0.01 M) was added to the solution as a bacterial inhibitor. In most phenanthrene experiments, the electrolyte solution was 0.1 M ionic strength solution containing 0.01 M CaCl~2Hz0and 0.0125 M NazB407.10 HzO as buffer at pH 8.0. A 0.005 M ionic strength solution of 0.001 M CaCl~2H20,0.001 M NaHC03, and 0.0004 M MgClz was used to mimic the composition of groundwater (38). A naphthalene solution of 0.041-1.137 mg/L was prepared prior to each adsorption experiment using the [lWInaphthalene stock solution and the electrolyte solution. Similarly, a phenanthrene solution of 0.006-0.279 mg/L was prepared using the [l4C1phenanthrene stock solution. Adsorption/Desorption Experiments. A typical batch reactor consisted of a glass vial of approximately 26 mL total volume (Wheaton) and capped with a Mininert valve. At the beginning of the adsorption experiment, 2 g of dried sediment was added to the vial before the addition of 25 mL of naphthalene solution. After being filled with sediment and solution, the vial typically contained a very small head space of less than 0.1 mL. The vial was set horizontally in a shaker bath and shaken at a low speed at room temperature. At the end of a desired incubation period, the soil was separated from solution by centrifugation at 300g (International Clinical Centrifuge, W. H. Curtin Co.) for 15 min, and the supernate was analyzed for solution-phase naphthalene concentration by liquid-scintillation counting (Beckman LS 3801). The centrifugation was able to settle all suspended particles greater than about 0.38pm and resulted a clear supernatant (39). After the adsorption experiments, desorption was induced by a successive replacement of 60 or 80% of supernate with electrolyte solution, and the vial was shaken. All dilution volumes were determined by weight for increased precision. The desorption times varied from 2 h to 137 days. From 2 to 20 successive desorption steps were employed in these experiments. Various adsorption/
desorption times, ionic strengths, and naphthalene concentrations were tested in separate experiments, and details of the procedures and results are listed in Tables 1and 2 for the adsorption experiments and in Table 3 for the desorption experiments. Solvent Extraction. At the end of adsorptionfdesorption experiments, the sediments were extracted with either methanol or methylene chloride for 24 h to determine the mass balance. For most of experiments, the supernate was removed by pipetting before the addition of methanol. A small amount of pore water was left in the compacted soil; this was about 4% (v/v). In methylene chloride extraction, the solvent was refluxed in a roundbottom flask containing soil for 24 h in a 45 "C water bath. In separate experiments, the amount of pollutants released from soil organic matters was measured by adding 0.5 N NaOH or 30% HzOz to dissolve or oxidize the organic matter. Data Analysis. In the adsorption experiments, the mass of naphthalene which disappeared from the solution phase at the end of the adsorption experiment was assumed to be adsorbed on sediments:
where qadaorbed is the mass of solute adsorbed on the sediments (pg/g), Cifitid is the initial concentration of solute (pg/mL), Cads is the solution-phase solute concentration at the end of the adsorption experiment (pgfmL), Vw is the volume of solution (mL),and W ,is the mass of sediment (g). The accuracy of this assumption was determined by mass balance analysis where the mass of contaminants present at the completion of the test was compared to the masses recovered by solvent extraction (see Table 1). A linear isotherm is assumed to model the adsorption of pollutants to sediment, with the following equilibrium Environ. Scl. Technol., Vol. 28, No. 5, 1994 861
Table 2. Summary of Phenanthrene Adsorption Experiments experimental no.a IP 2P 3P 4P 5P 6~ 7P 8P 9P 10P IlP 12P 13P 14P 15P 16P 17P 1 8 ~
soil fraction
organic carbon content (%)
adsorption time (day)
CfUd (rglmL)
Qadaorbd'
(x/mL)
(rglg)
K,d
bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk bulk 707-1000 pm 63-88 pm
