Mechanisms of Slow Sorption of Organic Chemicals to Natural Particles docx

Critical Review Mechanisms of Slow Sorption of Organic Chemicals to Natural Particles J O S E P H J . P I G N A T E L L O * A N D B A O S H A N X I N G Department of Soil and Water, The Connecticut Agricultural Experiment Station, P.O. Box 1106, New Haven, Connecticut 06504-1106 The use of equilibrium expressions for sorption to natural particles in fate and transport models is often invalid due to slow kinetics. This paper reviews recent research into the causes of slow sorption and desorption rates at the intraparticle level and how this phenomenon relates to contaminant transport, bio- availability, and remediation. Sorption kinetics are complex and poorly predictable at present. Diffusion limitations appear to play a major role. Contending mechanisms include diffusion through natural organic matter matrices and diffusion through intraparticle nanopores. These mechanisms probably operate si- multaneously, but the relative importance of each in a given system is indeterminate. Sorption shows anomalous behaviors that are presently not well explained by the simple diffusion models, including concentration dependence of the slow fraction, distributed rate constants, and kinetic hysteresis. Research is needed to determine whether adsorp- tion/desorption bond energies may play a role along with molecular diffusion in slow kinetics. The pos- sible existence of high-energy adsorption sites both within the internal matrix of organic matter and in nanopores is discussed. Sorption can be rate-limiting to biodegradation, bioavailablity, and subsurface transport of contaminants. Characterization of mech- anism is thus critical for fate and risk assessment. Studiesareneededtomeasuredesorption kinetics under digestive and respiratory conditions in receptor organisms. Conditions under which the constraint of slow desorption may be overcome are discussed, including the addition of biological or chemical agents, the application of heat, and the physical alteration of the soil. Introduction Sorptionto naturalsolids is an underlying processaffecting thetransport,degradation, andbiological activity of organic compounds in the environment. Although often regarded as instantaneous for modeling purposes, sorption may in fact require weeks to many months to reach equilibrium. It was not until the mid to late 1980s that serious study of sorption kineticsinsoils andsedimentsbegan, despite early circumstantial evidence going back to the 1960s that the natural degradation of certain pesticides in the field slowed or stopped after a while (1, 2). Sorption kinetics of contaminants on airborne particles has just recently received attention (3). Fate, transport, and risk assessment models all contain terms for sorption; therefore, an understanding of the dynamics of sorption is crucial to their success. Ignoring slow kinetics can lead to an underestimation of the true extent of sorption, false predictions about the mobility and bioavailability of contaminants, and perhaps the wrong choice of cleanup technology. Kinetics can also be an important mechanistic tool for understanding sorption itself. In this paper, we focus on updating our knowledge of the causes of slow sorption and desorption. In addition, we discuss its significance to bioavailability and the remediation of organic pollutants. Much of the research in this area has been carried out in batch systems where particles are suspended in a well-mixed aqueous solvent. Thus, we restrict discussion to phenomena occurring on the intraparticle scale, that is, within individual soil grains or within aggregates that are stable in water. We shall exclude transport-related nonequilibrium behavior (“physi- cal nonequilibrium”), which may also play an important role in nonideal solute transport in the field and in some experimental column systems. Physical nonequilibrium is due to slow exchange of solute between mobile and less mobile water, such as may exist between particles or between zones of different hydraulic conductivities in the soil column, and occurs for sorbing and nonsorbing molecules alike. It can give rise to transport behavior (plume spreading, “tailing” of the solute curve, etc.) that looks much like sorption nonequilibrium. It is irrelevant * Corresponding author telephone: (203) 789-7237;fax: (203)789- 7232; e-mail address: Soilwatr@yalvem.cis.yale.edu. 0013-936X/96/0930-0001$12.00/0  1995 American Chemical Society VOL. 30, NO. 1, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 1 to bioavailability per se, except that microbial populations and/or activity may vary within the flow regime. Recent papers discussing physical nonequilibrium are available (4-9). We shall also exclude chemisorption involving covalent bondsaswell as“bound residue” formation, which is defined as any organic carbon remaining after exhaustive extraction that results from degradation of the parent molecule. It is safe to say that the mechanisms governing sorption rates are not fully established. Thus, this paper is partly speculative. Slow Sorption and the Sorption Distribution Coefficient Research over the last decade or so has made it clear that (1) the solid-phase to solution-phase distribution coef- ficients(K d ) routinelyarenotmeasuredattrueequilibrium; (2) the use of equilibrium rather than kinetic expressions forsorptionin manyfate andeffectsmodelsisquestionable; and (3) the kinetics of sorption are complex and poorly predictable. Inmost cases, theuptakeor releaseoforganics bynatural particles is bimodal in that it occurs in fast and slow stages. The division between them is rather arbitrary, but in many cases it occurs at a few hours to a few days. Hereafter, the term slow will be used to describe the fraction sorbed or desorbed in the slow stage. Adjectives such as resistant, recalcitrant, rate-limiting, slowly reversible, and nonequi- librium are also used in the literature. The magnitude of the slowfraction is nottrivial, as many long-term studies testify. Some recent examples appear in Table 1. During uptake, the apparent sorption distribution coefficient (K d app ) can increase by 30% to as much as 10- fold between short contact (1-3 d) and long contact times. The values listed in Table 1 should not be construed as predictive nor necessarily representative. Data are sparse, and our level of understanding is insufficient to make predictions. During the slow uptake stage, experimentally observed changes in solution-phase concentration can be small over periods of many hours and are easily masked by random analytical errors. Consequently, it has been common in many routine sorption experiments to falsely conclude that the system has come to equilibrium after 1 or 2 days. Desorption likewise often reveals a major slow fraction (10-96%) following a comparatively rapid release. Histori- cally contaminated (aged) samples, where contact times may have been months or years, can be enriched in the slow fraction owing to partial dissipation or degradation of more labile fractions before collection. The slow fraction of some pesticides was found to increase with contact time in the environment (10). When thetotalcontaminantpresent must bedetermined by extractionssuch as in field samples or in spiked samples where uncertainlosses occurred during an experimentsthe choice of extraction conditions is important to ensure complete recovery of the analyte. Extraction methods are TABLE 1 Recent Examples of Observed Slow Sorption or Desorption in Natural Sorbents a Uptake contact period (d) long short approx ratio b K d app (long)/ K d app (short) slow fraction b,c ref PCE in aquifer sand material 10 1 3 0.67 33 TeCB in aquifer sand material 100 1 10 0.90 33 pyrene in lake sediments 180 3 2 0.50 107 phenanthrene in lake sediments 180 3 2 0.50 107 picloram in various soils 300 7 1.5-3.9 0.33-0.74 126 lindane in subsurface fine sand (corrected for abiotic hydrolysis) 167 4.2 4 0.74 38 atrazine in soil 22 1 up to ∼0.3 127 metolachlor in peat 30 1 1.4 0.22-0.33 d 55 metolachlor in soil 30 1 1.6 0.31-0.37 d 55 1,3-dichlorobenzene in peat 30 1 1.3 0.14-0.39 d 55 1,3-dichlorobenzene in soil 30 1 1.4 0.19-0.48 d 55 Release sparging or leaching time remaining slow fraction b ref PCB-contaminated river sediments 7-d continuous removal 0.17-0.45 28 TCE-contaminated subsoil seven 1-d washings or 24 000 column PV e 0.25-0.27 34 TCE-, PCE-, toluene-, xylene-contaminated soils 14 washings over 7 d 0.48-0.94 128 atrazine-contaminated soil 70-d leaching at 1 PV e /d 0.56 20 metolachlor-contaminated soil 70-d leaching at 1 PV e /d 0.59 20 naphthalene-contaminated soils 3-d gas purge 0.1-0.5 47 EDB-contaminated soil 10-d batch desorption 0.96 25 naphthalene-spiked soil (3-90 d contact) 3-d gas purge 0.1-0.2 47 simazine-spiked soil 35-d in the field 0.9 27 naphthalene-spiked soil (1-, 7-, 30-d contact) many 2-h to 7-d washings g0.6 15 phenanthrene-spiked soil (7-20-d contact) 10 washings over 178 d 0.62 15 TCE-spiked soil (2.5-, 5.5-, 15.5-mo contact) five 1-d washings 0.10, 0.25,0.45 34 PAHs on urban aerosols 28 d (130 m 3 of moist N 2 ) 0.4-0.6 3 atrazine on soil (4-, 12-, or 24-d contact) six 6-d batch desorptions 0.35-0.55 127 a PCE, tetrachloroethene; TeCB, 1,2,4,5-tetrachlorobenzene; picloram, 4-amino-3,5,6-trichloropicolinic acid; lindane, γ-1,2,3,4,5,6-hexachlorocy- clohexane; PCB, polychlorobiphenyl congeners; EDB, 1,2-dibromoethane; TCE, trichloroethene; atrazine, 2-chloro-4-ethylamino-6-isopropylamino- 1,3,5-triazine;metolachlor,2-chloro- N -[2-ethyl-6-methylphenyl]- N -[2-methoxyethyl]acetamide;simazine,2-chloro-4,6-bis(ethylamino)-1,3,5-triazine]. b Listed as estimates from graphs and tables in original work and may be rounded. c Slow fraction ) 1 - K d app (short)/ K d app (long). d Concentration dependent. e PV, column pore (void) volumes. 2 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 1, 1996 commonly validated with freshly spiked soil samples. Unfortunately, validation is seldom performed on aged samples that are enriched in resistant fractions. Hot extraction with water-miscible solvents has been shown to be superior for extracting resistant fractions compared to traditionalmethods like solvent-shake atroomtemperature, purge-and-trap, and Soxhlet techniques (11-15). It is well known that recovery by nonmiscible solvents (e.g., hexane) decreases with aging (16, 17). Supercritical CO 2 extraction has not been fully investigated; but some reports indicate that, even inthe presenceof organicsolvent modifiers which increase its solvation power, it is inferior to hot solvent for extracting resistant fractions (18, 19). In some studies, the slow fraction is likely to have been underestimated due to incomplete recovery. This can lead to erroneous conclu- sions when some process of interest is being measured against the mass of contaminant believed to be present. For example, one may deem that biodegradation is suc- cessful when actually loss of only the labile fraction has been evaluated. Since K d is time-dependent on a scale well beyond that of most laboratory sorption experiments, the true extent of sorption is known for just a few systems. Many reported K d values represent principally the fast component rather than overall sorption (20). Free energy correlations involv- ing K d are thus brought into question. For example, molecular structure-K d relationships rest on the assump- tion of equilibrium or at least that all compounds have attained the same fractional equilibrium. However, sorp- tion rates can depend greatly on molecular geometry and electronic properties. This is clearly evident in regard to diffusionthrough a viscous mediumsuchasorganic matter or a pore structure (see below). Moreover, Brusseau and co-workers (21, 22) showed that a mass transfer coefficient determined from soil column elution was inverse log- linearly related to the octanol-water partition coefficient for closely related compounds and that polarity in the molecule caused an additional decline in the mass transfer coefficient. Further research is needed to determine to what degree nonequilibrium can influence free energy relationships of sorption. In general, the sorption equilibrium assumption in fate andeffectsmodels isinvalid when the fate/transportprocess of interest occurs over comparable or shorter time scales thansorption. Given that, one canimaginemany processes that might bemore sensitivetokinetic thanthermodynamic sorption behavior; for example, uptake by an animal that comes into brief or intermittant contact with the soil. The equilibrium assumption has been found to fail in a growing number of cases. There are numerous examples of long- term persistence in soils of intrinsically biodegradable compounds even when other environmental factors are not limiting for microbial growth (2, 23-25). These are backed by a laboratory study showing that aging of the soil-contaminant mixture prior tothe additionof microbes reduced bioavailability (26) and by a field study showing that aging reduced herbicidal activity (27). Also, the fact that bioremediation of soil often levels off after an initial rapid decline [e.g., PCBs (28) and hydrocarbons (29)] is believed to be due mostly, if not solely, to the unavailability of a resistant fraction. Finally, nonequilibrium sorption affects the hydrody- namic transport of contaminants by causing asymmetrical concentration vs time (elution) curves. In relatively homogeneous soil columns, this asymmetry is exhibited by early breakthrough, a decrease in peak breakthrough concentration, breakthrough front tailing,and elution-front tailing (5); whereas, nonsorbing solutes like 3 H 2 O or Cl - typically show little or no evidence of asymmetry. In more heterogeneous media as exists in the field, the effect of nonequilibrium sorption on transport is less distinct. Vadose(30) and saturatedzone (4) studies reveal a decrease in velocity and aqueous-phase mass of the contaminant plume,relative to a nonsorbing tracer, withincreasing travel time or distance. While this is consistent with a time- dependent increase in K d app due to rate-limiting sorption, an interpretationis complicatedby permeability variations in the flow field (physical nonequilibrium) as well as variability in K d itself within the substrata (7, 8). Both of these can lead to tailing via plume spreading. The relative importance of sorption nonequilibrium and physical non- equilibrium is likely to depend greatly on the heterogeneity of the flow field and the type of particles that make it up. Mechanisms Possible Rate-Limiting Steps. The potential causesofslow sorption are activation energy of sorptive bonds and mass- transfer limitations (molecular diffusion). Sorption can occurby physical adsorption on a surfaceor by partitioning (dissolution) into a phase such as natural organic matter (NOM). The intermolecular interactions potentially avail- able to neutral organic compoundssvan der Waals (dis- persion), dipole-dipole, dipole-induced dipole, and hy- drogen bondingsare common to both adsorption and partitioning. In solution these forces are fleeting. For example, themeanlifetimeof theH 2 O‚‚‚NH 3 hydrogen bond is 2 × 10 -12 s (31). Adsorption to a flat, unhindered, and rigid surface is ordinarily unactivated or only slightly activated and so should be practically instantaneous (32). Desorption, however, is generally activated. The kinetic energy of desorption (E des *) is the sum of the thermody- namic energy of adsorption (Q)si.e., the depth of the potential energy wellsand the activation energy of adsorp- tion (E ad *) (32). A physisorbed molecule where E ad * ) 0 and Q e 40 kJ mol -1 will have a lifetime on the surface of e∼10 -6 s (32). For these reasons, most small compounds might be expected to adsorb and desorb practically instantaneously at the microscale. However, there may be situations in which E ad * or E des * is much greater. Large or longmoleculesthatcan interact simultaneously at multiple points can be more difficult to desorb. There may be steric hinderance to desorption or adsorptionsan ink bottle- shaped pore is an example. Lastly, there may be a cooperative change in the sorbent induced by the sorbate that makes Q larger, as occurs in substrate binding to enzymes. We must be open to these possibilities for pollutant molecules in highly heterogeneous systems like soil particles. It is noteworthy thatevensmall, weaklypolar molecules like halogenatedmethanes,ethanes,and ethenes exhibit slow sorption/desorption in soils (25, 33, 34). The thermodynamic driving force for their sorption is hydro- phobic expulsion from water, but their main interaction with the surface is only by dispersion and weak dipolar forces. Most researchers, nevertheless, attribute slow kinetics to some sortofdiffusion limitation. This is almost certainly true because sorbing molecules are subject to diffusive constraints throughout almost the entire sorption/desorp- tion time course because of the porous nature of particles. Diffusion is random movement under the influence of a VOL. 30, NO. 1, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 3 concentration gradient (35). Particles are porous by virtue of their aggregated nature and because the lattice of individual grains in the aggregate may be fractured. Figure 1 is a conceptualization of a soil or sediment particle aggregate showing possible diffusion processes. Toreachall sorptionsites,diffusing moleculesmusttraverse bulk liquid, the relatively stagnant liquid “film” extending from the solid surface (film diffusion), pores within the particle (pore diffusion),andpenetrablesolidphases (matrix diffusion). Diffusion coefficients of organic molecules can be expected to decrease along that same order, but except forbulkaqueousdiffusion,fewdataare availableforrelevant natural particle systems. The observed kinetics in any region of the sorption vs time curve will reflect one or more of these diffusive constraints, which may act in series or parallel. Themixing that takesplace in mostexperimentsensures that bulk liquid or vapor diffusion is not rate-limiting. Likewise, film diffusion is probably not rate-limiting. Film diffusion of inorganic ions is reduced or eliminated with vigorous mixing(36). Weberand Miller(37) and laterMiller and Pedit (38) concluded that in well-mixed batch systems film resistance of lindane and nitrobenzene on subsurface materials was insignificant compared to intraparticle dif- fusion, but may have been significant for nitrobenzene in columns (39). Film diffusion is potentially rate-limiting for the initial fast stage of sorption; but it is not likely to be important in the long-term phenomena we have been considering. This leaves pore diffusion and matrix diffusion as likely rate-limiting steps in slow processes. Diffusion in pores canoccur inpore liquids oralong porewallsurfaces. Liquid and surface diffusion may act concurrentlyand are difficult to distinguish (40, 41). A model of hydrophobic sorption to mineral surfaces (42) postulates that sorption occurs on or in “vicinal” watersthe interfacial region consisting of relatively ordered sorbed water moleculessrather than on the bare surface itself. If this model is correct, liquid and surface diffusion practically merge. Surface diffusion is expected toincreaseinrelativeimportance: (i)invery small pores wherefluids aremore ordered andviscous,andwhere the sorbate spends a greater percentage of time on the surface; (ii) at high surface concentrations. Surface dif- fusion was invoked for porous resins (43) and activated carbon (44, 45) because intraparticle transport appeared to be faster than could be accounted for by liquid diffusion. A surface diffusion model was used to simulate sorption- desorption of lindane with some success (38). However, it has been argued that surface diffusion is insignificant on soil particles because of the discontinuity of the adsorbing surface (33), if not the low mobility of the sorbate itself (46). Kinetic Behavior. Proposed mathematical kinetic mod- els include first-order, multiple first-order, Langmuir-type second-order (i.e., first-order each in solute and “site”), and various diffusion rate laws. The equations and their incorporation into the advection-dispersion model for solute transport are available in several good reviews (5, 6, 40). All except the diffusion models conceptualize specific “sites” to/from which molecules may sorb in a first-order fashion. Most sorption kinetic models fit the data better by including an instantaneous, nonkinetic fraction de- scribed by an equilibrium sorption constant . None of the models are perfect, although diffusion models are more successful than first-order models when they have been compared (20, 41). First-order kinetics are easier to apply to transport and degradation models because they do not require knowledge about particle geometry. Fit to a particular rate law does not by itself constitute proof of mechanism. Nonmechanistic modelshavebeen employed also. Pedit and Miller (41), on considering the inter- and intraparticle heterogeneity of soil, modeled the months- long uptake of diuron by a stochastic model, which treated sorbate concentration (K d ) and first-order rate constant as continuously distributed random variables. We call attention to three features of slow sorption kinetics that, if fully explained, could lead to a deeper understanding of the causesofslow sorption. First, a single rate constant often does not apply over the entire kinetic part of the curve (20, 46-48). In the elution of field-aged residues of atrazine and metolachlor from a soil column, a model with a single diffusion parameter underestimated desorption at early times and overestimated desorption at late times (20). Mass transfer coefficients obtained by modeling elution curves depend on the contaminant residence time in the columnsi.e., the flow rate (49). In desorption studies, plots of the logarithm of fraction remaining vs time tend to show a progressive decrease in slope,indicating greaterandgreaterresistance todesorption (47). Hence, desorption in natural particles seems to be, kinetically speaking, a continuum. On considering that soil may be a continuum of compartments ordered by their desorption rateconstants,Connaughton et al. (47)modeled the increasing desorption resistance of naphthalene by assuming that the rate constant is distributed according to a statistical Γ densityfunction,itself having two parameters. The intrinsic heterogeneity of soils on many levelsse.g., polydisperse primaryand secondaryparticles,awiderange FIGURE 1. Schematic of a soil particle aggregate showing the different diffusion processes. Natural “particles” are usually aggregates of smaller grains cemented together by organic or inorganic materials. Porosity is due to spaces between grains and fissures in individual grains. 4 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 1, 1996 of pore sizes, and spatial variations of mineral and organic components on the micro-scale, etc.sis fully compatible with continuous kinetics. An underlying problem in studying slow sorption is that we are never dealing with a homogeneous sorption/diffusion medium. Second, the slow fraction (S sl ) is inversely dependent, often markedly, on the initial applied concentration, C o (14,46, 50-52), meaningthat itassumes greater importance at lower concentration. Equilibrium considerations alone may partly explain this: when the sorption isotherm is nonlinearsthat is, when N in the Freundlich equation (C s ) K F C a N , where C s and C a are the sorbed and aqueous concentrations),isless than unitysintraparticleretardation will increaseastheconcentrationinsidetheparticledeclines (38, 46, 48). However, in some studies it appears that the concentration dependence is steeper than expected based on equilibrium nonlinearity. In studies of TCE vapor sorption to various porous particles at 100% relative humidity, Farrell and Reinhard (46) showed that the slow fraction remaining after N 2 gas desorption was highly concentration-dependent and not well simulated by con- sidering only equilibrium nonlinearity. In batch experi- ments of a soil containing 1.26% OC (50), an empirical nonlinear expression was used to relate “slow fraction” (amount remaining after desorption to infinite dilution for 5 d) to initial concentration (S sl ∝ C o n ). The exponent n was found to be 0.90 for PCE, 0.73 for 1,2-dibromo-3- chloropropane, and 0.49 for TCE. The isotherm of TCE in the same soil was linear (N ) 1.01) (53). While the Freundlich parameters were not measured for the other two compounds, experience shows (25, 53, 54) that such compounds give linear or slightly nonlinear isotherms (N > ∼0.9) in soils that have a substantial amount of NOM. Thus, for TCE at least, the concentration dependence of the slow fraction is greater than the fast fraction. In a study of metolachlor and 1,3-dichlorobenzene in two soils (55), N was greater for sorption of a fast fraction (1 d contact time) than a slower fraction (thedifference between 30 and 1 d contact times). This means that the slower fraction becomes increasingly dominant as the total concentration declines. Third, sorption is often kinetically hysteretic, meaning that the slow state appears to fill faster than it empties. Further research must be done to validate this. Many examples exist of apparent “irreversible” sorption of some fractionsor at least exceedingly long times to achieve desorptionsfollowing relatively short contact times (1, 5, 15, 56-59). Hysteresis may be caused by experimental artifacts or degradation (1, 5, 56). Also, to fairly assess hysteresis from the desorptive direction requires that samples be at true equilibrium. Kan and co-workers (15) sorbed naphthalene and phenanthrene to a sediment (0.27% NOM). While uptake appeared to reach equilibrium in a few days, successive desorption stepssusually lasting 1-7 d and totalling as long as 178 dsreleased less than 40% of chemical, even from samples sorbed for only 1 d (Table 1). Good mass balance was obtained upon soil extraction with CH 2 Cl 2 at 45 °C. Miller and Pedit (38) examined sorption of lindane to a subsurface soil corrected for dehydrohalogenation reactions. They found that an intraparticle diffusion model, whose parameters were obtained from uptake, could account for mostbutnot all of the hysteresis observed upon sequential desorption. We note that the sorbed concen- trations declined by only 2- or 3-fold after the three-step desorption, and the model fit seems to worsen with step. Hadfurther stepsbeen performedto uncover more resistant fractions,itis possiblethatevenless of the hysteresis would have been accounted for. Harmon and Roberts (48) found theeffectivediffusioncoefficient of PCEin aquifersediment to be 2-4 times smaller in the desorptive direction. They cautioned that the sorptive diffusion coefficients were obtained by others using a different technique. Inspection oftheir datarevealsthat thetail endof thedesorption curves tends to flatten out, indicating a substantial fraction of PCE (∼20% of initital) that is overpredicted by the model, i.e., desorbs at a much slower rate. The above three features of slow sorption suggest but do not prove a departure from regular Fickian diffusion. Fickian diffusion is symmetrical with respect to sorption and desorption, and the diffusion coefficient is concentra- tion-independent provided the sorbate does not alter the sorbent properties (35). Further careful experiments are needed to confirm whether sorption in soils truly deviates from Fickian diffusion. If it does, one implication is that the making/breaking of bonds may play a role along with molecular diffusion insorption/desorption rate limitations, even for classically “noninteracting” compounds like aromatic hydrocarbons and chlorinated solvents. The behaviorsabove are inlargemeasureasignature of sorption to sites having a distribution of energies. If interaction with an array of sites is responsible for sorption in the slow state the following might be expected: (i) a distribution of desorption rate constants corresponding to a distribution of activation energies; (ii) inverse concentration depen- dence of the slow fractionsat low applied concentration, the higher energy sites (which are more important relative tothe fast state) arepopulatedpreferentially; and(iii)kinetic hysteresis since the activation energy of desorption is normally greater than that of sorption from/to a specific site. We may better understand the meaning of these observations in the context of the two models that have been put forth as the most likely causes of slow sorption in natural particles: the organic matter diffusion model (OMD) and the sorption-retarded pore diffusion model (SRPD). They are shown pictorially in Figure 2 and are discussed below. Organic Matter Diffusion. The OMD model postulates diffusion through NOM solids as the rate-limiting step (5, 21). This is intuitively satisfying given the abundant thermodynamic evidence that partitioning (dissolution) in NOM is the primary mechanism of sorption when NOM and water are sufficiently abundant (54). NOM can exist as surface coatings or discreet particles. Supporting the OMDmechanismare the following: (1) inverse correlations between mass transfer parameters and NOM content (20, 28, 50, 60, 61); (2) organic cosolvents increase the rate in accord with their ability to ‘swell’ NOM (62); (3) inverse linear free energy correlations between rate constant and K d or the octanol-water partition coefficient K ow (3, 21, 22); and (4) a decrease in rate for polar molecules capable of hydrogen bonding to acceptor groups within NOM (22). Yet these results are also consistent with SRPD if the active sorbent material in pores is taken to be NOM coatings on pore walls. Moreover, OMD is at odds with the observation of slow sorption in zero or extremely low NOM materials (33, 46, 63). We may ask: Are diffusion length scales and diffusion coefficients (D) in natural particles consistent with NOM VOL. 30, NO. 1, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 5 as the diffusive medium? Desorption of resistant field- aged pesticides (EDB, atrazine, metolachlor) in soil show little particlesize dependencedowntotheclay-size fraction (20, 25), suggesting that the upper limit diffusion length scale is on the order of the clay particles (10 3 -10 2 nm). If a single effective diffusivity (D eff ) applies over this radial length, D eff would equal ∼10 -17 cm 2 /s or less. Desorption of PCBs from river sediments (28) also showed no particle size effects and indicated diffusion length scales of ∼30 nm, corresponding to D eff of 10 -20 -10 -21 cm 2 /s. The true dimensions of NOM are essentially unknown, but thick- nesses of 30-1000 nm are not unreasonable for NOM coatings or discreet NOM particles. In regard to D eff values, the obvious analogy to NOM is synthetic organic polymers. The polymer-phase concept of humics is replete in the literature. Diffusion in polymers occurs by either a place change mechanism, in which movement is accomplished by cooperative interchange of positionof polymersegmentsand thepenetrating molecule, or by a defect mechanism where the penetrant may jump between lattice defects, voids, pores, etc. (64). Polymers are said to have glassy (condensed, rigid) or rubbery (expanded, flexible) structures with respect to the order andcohesiveforces of thepolymerchains. Likewise, humic substances are described as having condensed and ex- panded regions (65). Choosing a polymer to model NOM is difficult because NOM in situ is expected to be highly variable in its properties, even within the same contiguous material. Furthermore, structure of and sorption to NOM can be strongly affected by soil minerals (66, 67). Attempts have been made to estimate the cohesive forces holding the humic polymer chains together in relation to their effects on the diffusivity and solubility of sorbate molecules (28). The true valuesswere it possible to determine themsare likely to cover a wide range. ReportedD values in polymers at 25-30 °C for a molecule like CCl 4 having a diameter of 0.55 nm range over many orders of magnitude, from 10 -7 cm 2 /s in rubbery polymers (polyethylene) to 10 -17 cm 2 /s in glassy polymers (polyvinyl chloride) (64, 68). Diffusivity is sensitive to the size and shape of the penetrant, much more so for glassy than rubbery polymers. One might expect a molecule to experience large changes in diffusivity as it moves between expanded and condensed regions of NOM. Accordingly, Carroll et al. (28) suggest that the bimodal desorption vs time curves of PCBs from sediments aredue todesorption fromthesetwo typesof phases. Future work is needed on determining organic compound diffu- sivities in NOM particles and on finding appropriate polymer models. Diffusion kinetics in polymers is widely variable de- pending on polymer structure, particle size distribution, diffusant structure, diffusant concentration, temperature, andthe history ofexposure (64,69, 70). Mixturesof polymer sphere sizes can lead to bimodal diffusion curves (71, 72). Since the diffusion rate is inversely related to the square of the radius, the proportion of fast and slow phases of the uptake or release curve depends on the size distribution. Obviously, the dimensions of NOM in a given soil will be truly diverse. Relatively high diffusant concentrations can cause polymer swelling or crazing as the diffusant front advances (69). These changes affect both the compound’s solubility (partition coefficient) and diffusivity, which in turn dictate the shape of the kinetic curve. Bimodal curves can result. Pollutant concentrations in the environment may some- times be high enough (e.g., a chemical spill) to swell or soften NOM. Cosolvents can do the same thing. Methanol cosolvent increased the desorption rate constant of diuron and several PAHs (62). Usually though, we are dealing with dilute contaminant and no cosolvent. Sorption under dilute conditions in rubbery polymers generally is linear and obeys Fick’s lawsthat is, D is concentration-independent, and mass transfer is symmetrical with respect to the forward and reverse directions and proportional to the square root of time (64, 69, 73). Sorption in glassy polymers on the other hand is anomalous in that it is typified by nonlinear (N < 1) isotherms, concentration-dependent D, and a tendency toward bimodal kinetics and sorption-desorption hyster- esis (64,68,70, 72). This isreminiscentof sorption/diffusion behavior of many compounds in soils. Anomalous behavior in glassy polymers has been attributed to dual-mode sorption. This was first proposed FIGURE 2. Schematic of two models for slow sorption. (a) Organic matter diffusion (OMD),illustrating diffusion through a rubberyphase A,diffusionthrough a morecondensedglassyphaseB, and ad sorption in a “Langmuir site” C (see text). (b) Sorption-retarded pore diffusion (SRPD). Retardation by rapid-reversible sorption to pore walls, and “enhanced adsorption” in pores of very small diameter due to interaction with more than one surface. 6 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 1, 1996 for gaseous molecules like CO 2 and CH 4 (70) and later for small organic molecules (68, 71, 72). Dual-mode sorption is the sum of (a) normal linear partitioning taking place in the bulk of the polymer and (b) a hole-filling mechanism in which the incoming molecules undergo Langmuir-like adsorption in voids internal to the polymer matrix. The latter is the cause of isothermnonlinearity and non-Fickian tendencies. Linear sorption can be restored by conversion of the glassy state to the rubbery state by increasing the temperature above the glass transition point (T g ) or by softening with organic solvents. The exact nature of the voids is presently unknown. Solid-state 31 P-nuclear mag- netic resonance spectroscopy showed that mobile and immobile sorbed forms of tri-n-butyl phosphate exist in glassy polystyrene (74). The relative mobility of the immobile forms appeared to span a wide range. Dual-mode sorption in NOM may rationalize qualita- tively some behaviors of contaminants in natural parti- clessnamely, nonlinear isotherms, competitive sorption, and kinetichysteresis. Isotherms areoftennonlinear when a sufficiently wide solute concentration range is used (37, 39,55, 75-77). Examplesinclude hydrophobiccompounds in a peat soil composed almost entirely (93%) of NOM (55). It might be expected that isotherms would linearize at high concentrations as the adsorption sites became filled, but this would depend on how the sites were distributed in energy. Investigatorshave alsoshowncompetitivesorption between nonpolar compounds in suspensions of soils (53, 76), the mentioned peat (53), and humic-coated clay (66). Competitive sorption clearly indicates some measure of site specificity (53, 76). As shown in Figure 2, we may envision NOM as a bulk partition medium consisting of rubbery (A) and glassy (B) regions. Dispersed in the glassy regions are adsorption sites(C)ofvarious energies, analogous to the voidsofglassy polymers. In agreement with the dual-mode model, phenanthrene isotherms became more linear with increas- ing temperature in soil and shale samples where NOM was believed to be the predominant sorbent (75). This is consistent with a transition to a more rubbery state. The nature of the adsorption sites is speculative. They could be some type of inclusion complex between the guest pollutant molecule and host subunit(s) on the NOM macromolecule. Soil humic acid has condensed polyaro- matic regions, even after extraction and reconstitution to a particulate form (78, 79). It has been suggested that polyaromatic structures provide adsorption sites (75). Although it is far from certain at this time that the dual- mode mechanism plays a role in slow kinetics, the aforementioned results of ours (55), showing a decrease in the Freundlich exponent N with time in NOM, are at least consistent with it. Theexistence of high-energyadsorption sites could account for kinetic hysteresis. It might be expected that such sites would fill faster than they would empty. Thus, it is plausible that desorption becomes at some point rate limited by release from these sites, while sorption is principally rate limited by diffusion through bulk NOM. The nonlinear relationship between S sl and C o discussed above is also plausibly attributed to the presence of sites. Sorption-Retarded Pore Diffusion. The SRPD model (Figure 2) postulates the rate-limiting process to be mo- lecular diffusion in pore water that is retarded, chromato- graphic-like, by local sorption on pore walls (80) (Figure 2). Walls may or may not be composed of NOM. Assumptions by most modelers are that local sorption is instantaneous, particles are uniformly porous, and sorption parameters K d and D eff in the pore are constant. According to the SRPD model, rates are expected to be inversely dependent on the square of the particle radius, on the tortuosity of pores (bendingandtwisting,interconnectivity,presence of dead- end pores), on the constrictivity (steric hindrance) in the pores, and on the K d . The inverse dependence on K d does not distinguish SRPD from OMD. For natural particles, observations that point to SRPD include faster rates after particle pulverization, which reduces pore pathlength (25, 33,50), and afteracidification, which was suggested to disagregate grains by dissolving the inorganic oxide cements that hold the aggregates together (50). Correlation of rate with particle size is only qualitative at best. In one case where a rough correlation was found (80), experiments took place over a few hours at most. In another case (33) where coarse aquifer sand particles equilibrated PCE and TeCB generally faster than fine particles, the particles were calcite-cemented ag- gregates thathad considerableinternalporosityand surface area (81). In many systems, the particle size dependence of desorption is altogether absent (20, 25, 28, 46, 82). For example, desorption of field-aged pesticides in soil (20, 25) and PCBs in river sediments (28) was not related to the nominal particle radius down to the clay size fraction, suggesting that the length scale of diffusion is proabably less than 100 nm. The absence of size dependence might be rationalized by assuming that most of the porosity exists in an outer shell that is of similar thickness among size fractions. Anotherpossiblereasonisthat sorption capacity may not be uniformly distributed within the aggregate. Ball and Roberts (33), for example, found that K d of PCE and TeCB varied markedly among different size fractions and a magnetically separated fraction of an aquifer material. But in a study of desorption of TCE from silica with monodisperse particle sizes and narrow pore size distribu- tions, investigators found no particle size dependence (46). Tortuosity and constrictivity are difficult to evaluate. Both areexpectedto varyinverselywith pore size. However, in a silica pore, diameters ranging from 6 to 30 nm had little effect on TCE desorption (46). It is possible that this effect shows up only in pores that are smaller than 6 nm. Theanalyticaltoolsformeasuring nanoporecharacteristics in natural materials are undeveloped, and the theoretical foundations aretoo weak to incorporate them intothe SRPD model (33, 46, 48). Thus, research is needed on character- izing the geometry and spatial distribution of pores. The kinetic continuum discussed earlier could be rationalized by a heterogeneous pore structure in which there is a distribution of diffusivities within the particle. Thus, we may envision pores that fill and empty quickly, along with those that do so slowly. A potentially important influence on constrictivity in the pore is the viscosity of water. Polar minerals have one or more layers of water strongly sorbed on their surfaces (83). Viscosity measurements of colloidal silica particles in water indicate there is a monolayer more or less immobilized on the surface (84). The water contained in a pore of a few Angstroms in diameter may be ice-like and therefore greatly restrict solute diffusion. Molecular mod- eling can potentially give insight on the structure of water in nanopores. Thelong-term desorptionofEDBfromsoil towatercould not be modeled by SRPD without invoking enormous VOL. 30, NO. 1, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 7 tortuosity or constriction in pores (25). Likewise, D eff ’s for desorption of alkanes and PAHs from urban atmospheric particles were 10 6 times smaller than expected for sorption retarded gaseous diffusion in pores (3). One rationale is affordedby rejecting theSRPDassumptionof instantaneous local equilibrium in the pore. For instance, a PCB required hours to desorb from dissolved commercial humic acid (85). From the standpoint of solutes, dissolved humic acid macromolecules may well represent the smallest or most penetrablehumic materialsexisting on wall surfaces. Also, substituted benzenes required hours to desorb from surfaces of alkyl-modified, nonporous silica gel particles (86). Such slow stepwise desorption rates could strongly retard transport through a pore compared to the instan- taneous case. The mathematics of diffusion in systems containing one diffusive medium in another have been discussed (35). An alternative explanation for the extraordinarily small D eff has been offerred (46, 87). According to this enhanced adsorption hypothesis, asthepore sizedecreases toroughly theadsorbate diameter, the calculatedinteraction potential increases up to 5-fold compared to the single surface case owing to multipoint interaction of the adsorbate with pore walls (Figure 2). Furthermore, since the pore volume is small compared to the wall surface area, the adsorbate spends less and less time in pore solution, ultimately being restricted to diffusion along the surface, which may be intrinsically slower. Farrell and Reinhard (46, 87) equili- brated TCE vapors with a column of porous silica particles at100% relativehumiditywhereall microporesare expected to be filled with water. Desorption with a stream of humidified N 2 proceeded in two distinct phases and was incomplete after purging times lasting weeks. The silicas behavedsimilarly tonatural sorbents inthat(1) thediffusion length scale was not the nominal particle radius and (2) the slow fraction was less than linearly related to initial TCE concentration, which was attributed to a limited sorption capacity in high-energy micropore sites. The authors suggest that microporosity gives rise to both isotherm nonlinearitysindicative of a distribution of site ener- giessand slow desorption. Here again, we see the con- sequences of energetic heterogeneity in the sorbent that was referred to earlier in a general sense and in the context of OMD. Cautioniscertainlycalledfor in interpreting these experiments. Condensation of TCE in pores during the equilibration period cannot be ruled out since vapor concentrations were close to saturation. Removal of TCE in a condensed phase from a pore may be slower. In apparent contradiction of enhanced adsorption is the fact that diffusion of small molecules through the micropores (5-7 Å) of synthetic aluminosilicate zeolites is remarkably faststhe time to reach equilibrium of small molecules like hexane (88) and TCE (89) in micron-size particles being on the order of 10 2 min (D ∼ 10 -12 cm 2 /s). A form of pore diffusion that deserves more attention is clay interlayer diffusion. Hydrated metal ion-exchanged clays (e.g., with Ca 2+ ) do not extensively sorb hydrophobic molecules, but neither are such compounds excluded from hydrated interlayer spaces. Na-montmorillonite in water exhibited uptake of TCE lasting over 25 d (90). Desorption of atrazine from some Ca 2+ -smectites revealed formation of a tightly bound fraction (91). Smectites exchanged with tetraorganoammonium cations haveamuch higher affinity forhydrophobiccompounds (ref92and references therein), but their kinetics have not been studied. Some evidence suggests that clay interlayers are not important. Mont- morillonite formed a much smaller fraction of slowly released TCE than either silica or microporous glass beads (46). Steinberg et al. (25) observed the lowest field residues of EDB in the clay size fraction. ConcludingRemarksRegarding Mechanism. Itisquite likely that both OMD and SRPD mechanisms operate in the environment, often probably together in the same particle. OMD may predominate in soils that are high in NOMand low inaggregation,while SRPDmay predominate in soils where the opposite conditions exist. But this has not been established. Slow desorption from an organic- free silica, a substance so closely related to soil minerals, is strong evidence that the mineral fraction is important. Resolving the individual contributions of OMD and SRPD in natural materials constitutes a challenge to future investigators. We have seen that both mechanisms offer the potential for high-energy adsorption sites to play a role. These sites may be more rate-limiting in the desorptive direction than the sorptive direction. Further research is critical in this area. Evidence indicates that a decrease in the rate constant occurswith increasing molecular size and hydrophobicity. However, this is consistent with all of the mechanisms discussed. Significance of Slow Sorption Mechanism to Bioavailability and Remediation The bioavailability of chemicals in soil to microbes, plants, and animals is important from the perspective of reme- diation and risk assessment. Ex situ or in situ cleanup of soil requires mass transport of contaminants through the materials, which in turn depends on sorption kinetics. Microbes take up substrates far more readily from the fluid than the sorbed states (89, 93-96). Thus, it is no surprise that aged chemicals are resistant to degradation compared to freshly added chemicals (25, 27, 97, 98) and that degradation of freshly added chemicals often tails off to leave a resistant fraction (26, 98-100). Bioavailability has been called a major limitation to complete bioreme- diation of contaminated soils (29, 101). The soil- contaminant-degrader system is dynamic and interde- pendent. A mechanistic-basedbiodegradationmodelmust be built on the mechanism(s) governing sorption/desorp- tion, in addition to the biological mechanisms governing cell growth and substrate utilization in the matrix. A number of groups are now developing sorption-degrada- tion kinetic models (26, 102-106). Both diffusion and two- box (equilibrium and first-order kinetic compartments) sorption concepts have been explored. The bioavailability of pollutants to wildlife and humans is also an area of critical importance. Pollutants can be taken up in pore water, by dermal contact, by particle ingestion, or by particle inhalation. The dynamics of sorption are not currently incorporated into exposure and risk models for organics. Availability in most cases is assumed to be 100% (107). Recently, the following have been demonstrated: (1) the time between spiking and testing affects bioavailability (2, 108); (2) the kinetics of desorption control bioaccumulation of historical contami- nation (e.g., PAHs in benthic animals; 109); and (3) historically contaminated soils are less toxic and/or lead to lower body burdens than equivalent amounts of spiked soils (110, 111). In order to model bioavailability, it is crucial that we understand sorption kinetics and the factors that influence 8 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 1, 1996 rates under the conditions of exposure. Take particle ingestion, for instance. The intestines of warm-blooded animals are often at higher temperature than the soil being ingested. Molecular diffusion through a viscous medium like NOM and desorption from a surface are activated processes and hence temperature sensitive. For example, theapparentactivation enthalpy for desorption of historical residues of EDB from soil into water was 66 kJ/mol, corresponding to a 7-fold rate increase from 25 to 40 °C (25). The application of heat increased the rate of desorp- tion of PCBs from river sediment and reduced the resistant fraction (28). Also, there is evidence that pH is important. Acidification of a soil suspension to pH <∼2 accelerated desorption of the slow-desorbing fraction of several ha- logenated aliphatic hydrocarbons. The amounts desorbed in 1 h ranged from 13 to 80% of the slow fraction (50), many times more than the control at natural pH. The human stomach can be highly acidic (pH 1.5-2) at times (112). Soilingested by birdsissubjected to grindinginthegizzard, which may release slow-desorbing contaminants if pore diffusion isimportant. Furtherworkis needed to determine what physiological conditions in both the digestive and repiratory tracts may impact desorption, and further work is needed on the design of experiments to accurately simulate such conditions in the laboratory. Vapor and water extraction methods (pump-and-treat), which are widely used in remediation, are limited in part by physical nonequilibrium and sorption nonequilibrium (113-115). These processes both cause tailing of the contaminant plume, whichincreases the time invested and thevolumeofsparge air or waterneeded to achieve cleanup (113, 115-117). Moreover, they act to resume contamina- tion if pumping is ceased before all the contaminant is removed (rebound) (118-120). Ways of experimentally separating out the contributions of physical and sorption nonequilibrium must be sought. Experience is proving that the constraint of slow desorption has to be overcome to achieve complete remediation (29). We may consider the following conceiv- able approaches to promoting desorption from the slow state: (1) addition of biological agents capable of reaching remote molecules; (2) application of heat; (3) addition of chemical additives that displace the contaminant or alter the soil structure; and (4) physical methods that alter the soil structure. Since cells, being g0.2 µm, are too large to fit in nanopores orwithin theNOM matrix,it wouldseem unlikely that strainsexist which candirectlyattackremote molecules. Guerin and Boyd (106) isolated a Pseudomonad that, compared to another degrader, appeared to enhance the desorption of naphthalene by providing a steep concentra- tion gradient at the particle surface. Some organisms, especiallyfungi,metabolizecontaminantswith extracellular enzymes. Enzymes may also be added to soil to remediate it. However, enzymes are many times larger than con- taminant molecules and probably diffuse far more slowly, if at all, through micropores or NOM to reach resistant molecules. Moreover, sorption of enzymes may reduce their activity (121). As mentioned, molecular diffusion through NOM and desorption from high-energy sites are expected to be strongly temperature dependent. Thermal desorption is already in use in various remediation technologies for contaminants of sufficient volatility. In batch application, the soil is heated to temperatures ranging from 200 to 500 °C in a primary chamber, and the vapors are combusted inasecondarychamber(122-124). Steam stripping (aform of soil vapor extraction) can remove semivolatiles from the vadose zone (125). Bioremediation in a compost mode where temperatures reach 60 °C or more should prove advantageous. The success of these methods requires a fundamental understanding of kinetics. Research into sorption kinetics in regard to steam stripping has been initiated (125). Alteration of the soil chemistry is another approach that should be considered further based on preliminarystudies. As mentioned, acidification promoted desorption (50), but further work is needed to determine its scope and prac- ticality. The use of surfactants targeted specifically to removal of slow fractions has not yet been adequately addressed in theliterature. Tobe effective, surfactants must penetrate the intraparticle matrix (nanopores or NOM) to either (i) solubilize the contaminant by micellization or (ii) alter the intraparticle properties of the sorbent in such a way as to promote desorption. The addition of surfactants gave mixed results in stimulating biodegradation (95, 126). The use of organic cosolvents is a promising approach because cosolvents can increase desorption both thermo- dynamically (by enhancing solubility) and kinetically (by softening NOM) (62). Supercritical carbon dioxide extrac- tion has been proposed for large-scale cleanup (19). It probably would require up to 10% by weight of a polar organic cosolvent to increase its solvation power. This is an example where an understanding of sorption kinetics would prove beneficial. Physical manipulation of the soil such as grinding is known to be partially effective (25, 33, 50, 63) but would likely be impractical on a large scale. Summary Remarks Sorption and especially desorption in natural particles can be exceedingly slow. The rate-limiting nature of sorption has widespread implications but is poorly understood and predicted. The importance of it is appreciated by con- sidering that if sorption occurs on time scales of months or longer, true equilibrium may exist in only limited environments. It is hoped that researchers who deal in fate and transport of contaminants are by now more aware of the phenomenon itself as well as the potential for misinterpretation that can result if kinetics are ignored. Slow sorption has made complete remediation difficult. However, there have been legitimate questions raised by some (2, 29, 107) about whether we even need to be concerned about residues that desorb so slowly and are apparently largely bio-unavailable. At a minimum, it is critical that we understand the factors that govern their release. Sorption kinetics are extremely important in modeling the transport of contaminants in the subsurface. Understanding the causes of slow sorption/desorption hasbeen hamperedbythe heterogeneity ofnaturalparticles as a sorptive and diffusive medium. It is no wonder then that rate parameters seem to depend in a complex way on the soil, history of exposure, and even position along the uptake vs time curve. But we can be confident that kinetics studies will lead to a deeper understanding of sorption mechanism itself. Future research should focus not only on understanding andpredictingthe rates ofslow sorption/ desorption but also on overcoming the constraints of slow desorption for remediation purposes. VOL. 30, NO. 1, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 9 Acknowledgments We thank the U.S. Department of Agriculture National Research Initiative (Water Quality) for support and the reviewers of the manuscript for their suggestions. Literature Cited (1) Pignatello, J. J.In Reactions and Movementof Organic Chemicals in Soil; Sawhney, B. L., Brown, K., Eds.; Soil Science Society of America: Madison, WI, 1989; Chapter 3, pp 45-97. (2) Alexander, M. In Draft Report Environmentally Acceptable Endpoints in Soil; Gas Research Institute, Environment & Safety Research Group 1995; Chapter 1. (3) Rounds, S. A.; Tiffany, B. A.; Pankow, J. F. Environ Sci. Technol. 1993, 27, 366-377. (4) Roberts, P. V.; Goltz, M. N.; Mackay, D. M. Water Resour. Res. 1986, 22, 2047. (5) Brusseau, M. L.; Rao, P. S. C. Crit. Rev. Environ. Control 1989, 19, 33. (6) van Genuchten, M. T.; Wagenet, R. J. Soil Sci. Soc. Am. J. 1989, 53, 1303. (7) Goltz, M. N.; Roberts, P. V. J. Contam. Hydrol. 1988, 3, 37-63. (8) Velocchi, A. J. Water Resour. Res. 1989, 25, 273-279. (9) Gaber, H. M.; Inskeep, W. P.; Comfort, S. D.; Wraith, J. M. Soil Sci. Soc. Am. J. 1995, 59, 60-67. (10) Pignatello, J. J.; Huang, L. Q. J. Environ. Qual. 1991, 20, 222. (11) Huang, L. Q.; Pignatello, J. J. J. Assoc. Off. Anal. Chem. 1990, 73, 443. (12) Sawhney, B. L.; Pignatello, J. J.; Steinberg, S. M. J. Environ. Qual. 1988, 17, 149. (13) Pignatello, J. J. Environ. Toxicol. Chem. 1990, 9, 1107. (14) Steinberg, S. Chemosphere 1992, 24, 1301-1315. (15) Kan, A. T.; Fu, G.; Tomson, M. B. Environ. Sci. Technol. 1994, 28, 859-867. (16) Karickhoff, S. W. In Contaminants and Sediments, Vol. 2; Baker, R. A., Ed.; Ann Arbor Science: Ann Arbor, MI, 1980; pp 193-205. (17) Eischenback, A.; Kaestner, M.; Bierl, R.; Schaefer, G.; Mahro, B. Chemosphere 1994, 28, 683-692. (18) Koskinen, W. C.; Cheng, H. H.; Jarvis, L. E. J.; Sorenson, B. A. Int. J. Environ. Anal. Chem. 1994, 58, 379-385. (19) Laitinen, A.; Michaux, A.; Aaltonen, O. Environ. Technol. 1994, 15, 715-727. (20) Pignatello, J. J.; Ferrandino, F. J.; Huang, L. Q. Environ. Sci. Technol. 1993, 27, 1563-1571. (21) Brusseau, M. L.; Jessup, R. E.; Rao, P. S. C. Environ Sci. Technol. 1991, 25, 134. (22) Brusseau, M. L.; Rao, P. S. C. Environ. Sci. Technol. 1991, 25, 1501. (23) Pignatello, J. J.; Frink, C. R.; Marin, P. A.; Droste, E. X. J. Contam. Hydrol. 1990, 5, 195. (24) Alexander, M. Biodegradation and Bioremediation; Academic Press: New York, 1994; Chapters 10 and 11. (25) Steinberg, S. M.; Pignatello, J. J.; Sawhney, B. L. Environ. Sci. Technol. 1987, 21, 1201. (26) Hatzinger, P. B.; Alexander, M. Environ. Sci. Technol. 1995, 29, 537-545. (27) Scribner, S. L.; Benzing, T. R.; Sun, S.; Boyd, S. J. Environ. Qual. 1992, 21, 115. (28) Carroll, K. M.; Harkness, M. R.; Bracco, A. A.; Balcarcel, R. R. Environ. Sci. Technol. 1994, 28, 253-258. (29) Loehr, R. C.; Webster, M. T. In Draft Report Environmentally Acceptable Endpoints in Soil; Gas Research Institute, Environ- ment & Safety Research Group: 1995; Chapter 2. (30) Jaynes, D. B. Soil Sci. Soc. Am. J. 1991, 55, 658-664. (31) March, J. Advanced Organic Chemistry, 3rd ed.; John Wiley & Sons: New York, 1985; Chapter 3. (32) Adamson, A. W. The Physical Chemistry of Surfaces; John Wiley & Sons: New York, 1976. (33) Ball, W. P.; Roberts, P. V. Environ. Sci. Technol. 1991, 25, 1237. (34) Pavlostathis, S. G.; Jaglal, K., J. Environ. Sci. Technol. 1991, 25, 274. (35) Crank, J. The Mathematics of Diffusion; Oxford University Press: Oxford, United Kingdom, 1975. (36) Sparks, D. L. Kinetics of Soil Chemical Processes; Academic Press: San Diego, CA, 1989. (37) Weber, W. J.; Miller, C. T. Water Res. 1988, 22, 457-464. (38) Miller, C. T.; Pedit, J. A. Environ. Sci. Technol. 1992, 26, 1417. (39) Miller, C. T.; Weber, W. J. Water Res. 1988, 22, 465-474. (40) Weber, W. J. J.; McGinley, P. M.; Katz, L. E. Water Res. 1991, 25, 499. (41) Pedit, J. A.; Miller, C. T. Environ. Sci. Technol. 1994, 28, 2094- 2104. (42) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental OrganicChemistry:JohnWiley &Sons: NewYork, 1993; Chapter 11. (43) Komiyama, H.; Smith, J. M. Am. Inst. Chem. Eng. J. 1974, 20, 728-734. (44) Crittenden, J. C.; Weber, W. J. Jr. J. Environ. Eng. Div. (Am. Soc. Civ. Eng.) 1978, 104, 185-197. (45) Critenden, J. C.; Hand, D. W.; Arora, H.; Lykins, B. W. J. Am. Water Works Assoc. 1987, 79, 74-84. (46) Farrell, J.; Reinhard, M. Environ. Sci. Technol. 1994, 28, 63-72. (47) Connaughton, D. F.; Stedinger, J. R.; Lion, L. W.; Shuler, M. L. Environ. Sci. Technol. 1993, 27, 2397-2403. (48) Harmon, T. C.; Roberts, P. V. Environ Sci. Technol. 1994, 28, 1650-1660. (49) Brusseau, M. L. J. Contam. Hydrol. 1992, 9, 353. (50) Pignatello, J. J. Environ. Toxicol. Chem. 1990, 9, 1117. (51) Grathwohl,P.; Reinhard, M.Environ Sci. Technol.1993,27, 2360- 2366. (52) Burris, D. R.; Antworth, C. P.; Stauffer, T. B. Environ. Toxicol. Chem. 1991, 10, 433. (53) Pignatello, J. J. In Organic substances and Sediments in Water Vol. 1; Baker, R. A., Ed.; Lewis Publishers: Chelsea, MI, 1991; pp 291-307. (54) Chiou, C. T. In Reactions and Movement of Organic Chemicals in Soil; Sawhney, B. L., Brown, K., Eds.; Soil Science Society of America, Madison, WI, 1989; Chapter 1, pp 1-29. (55) Xing, B.; Pignatello, J. J. ACS Environ. Chem. Div. Proc. 1995, 35 (1), 432-435. (56) Rao,P.S. C.; Davidson,J.M. In EnvironmentalImpactofNonpoint Source Pollution; Overcash, M. R., Davidson, J. M., Eds.; Ann Arbor Science Publishers: Ann Arbor, MI, 1980; pp 23-67. (57) Pignatello, J. J. Environ. Toxicol. Chem. 1991, 10, 1399. (58) Di Toro, D. M.; Horzempa, L. M. Environ Sci. Technol. 1982, 16, 594. (59) Fu, G.; Kan, A. T.; Tomson, M. Environ. Toxicol. Chem. 1994, 13, 1559-1567. (60) Nkedi-Kizza, P.; Brusseau, M. L.; Rao, P. S. C.; Hornsby, A. G. Environ. Sci. Technol. 1989, 23, 814. (61) Karickhoff, S. W.; Morris, K. R. Environ. Toxicol. Chem. 1985, 4, 469. (62) Brusseau, M. L.; Wood, A. L.; Rao, P. S. C. Environ. Sci. Technol. 1991, 25, 903. (63) Ball, W. P.; Roberts, P. V. Environ. Sci. Technol. 1991, 25, 1223- 1237. (64) Rogers, C. E. In The Physics and Chemistry of the Organic Solid State; Interscience Publishers: New York, 1965; Vol. 2, pp 509- 634. (65) Hayes, M. H. B.; Himes, F. L. In Interaction of Soil Minerals with Natural Organics and Microbes; Huang, P. M., Schnitzer, M., Eds.; Soil Science Society of America: Madison, WI, 1986; p 103. (66) Murphy, E. M.;Zachara,J. M.;Smith,S. C.;Phillips, J. L.; Wietsma, T. W. Environ. Sci. Technol. 1994, 28, 1291-1299. (67) Laird,D. A.; Yen, P.Y.;Koskinen,W. C.; Steinheimer, T.R.;Dowdy, R. H. Environ. Sci. Technol. 1994, 28, 1054-1061. (68) Berens, A. R. Makromol. Chem. Macromol. Symp. 1989, 29, 95- 108. (69) Koenig, J. L. In Physical Properties of Polymers, 2nd ed.; Mark J. E., et al., Eds.; American Chemical Society: Washington, DC, 1993; Chapter 6, pp 263-312. (70) Vieth, W. R. Diffusion In and Through Polymers; Hanser Verlag: Munich, 1991 (distributed in U.S. and Canada by Oxford University Press, New York). (71) Berens, A. R. Polymer 1977, 18, 697-704. (72) Berens, A. R. J. Membr. Sci. 1978, 3, 247-264. (73) Reynolds, G. W.; Hoff, J. T.; Gillham, R. W. Environ. Sci. Technol. 1990, 24, 135-142. (74) Toscano, P. J.; Frisch, H. L. J. Polym. Sci. 1991, 29, 1219-1221. (75) Young, T. M.; Weber, W. J., Jr. Environ. Sci. Technol. 1995, 29, 92-97. (76) McGinley,P. M.; Katz,L. E.; Weber,W. J., Jr.Environ Sci. Technol. 1993, 27, 1524-1531. (77) Spurlock, F. C.; Huang, K.; Van Genuchten, M. Th. Environ. Sci. Technol. 1995, 29, 1000-1007. (78) Chen, Z.; Pawluk, S. Geoderma 1995, 65, 173-193. (79) Schnitzer, M.; Kodama, H.; Ripmeester, J. A. Soil Sci. Soc. Am J. 1991, 55×e2 745-750. (80) Wu, S.; Gschwend, P. M. Environ Sci. Technol. 1986, 20, 717. (81) Ball, W.P.; Buehler, C. H.;Harmon, T. C.; Mackay,D. M.; Roberts, P. V. J. Contam. Hydrol. 1990, 5, 253-295. (82) Novak, J. M.; Moorman, T. B.; Karlen, D. L. J. Agric. Food Chem. 1994, 42, 1809-1812. (83) Drost-Hansen, W. J. Colloid Interface Sci. 1977, 58, 251. (84) Dalton, R. L.; Iler, R. K. J. Phys. Chem. 1956, 60, 955 10 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 1, 1996 [...]... C.; Zehnder, A J B Environ Sci Technol 1990, 24, 1349-1354 (103) Rao, P S C.; Bellin, C A.; Brusseau, M L In Sorption and Degradation of Pesticides and Organic Chemicals in Soil; Linn D M., et al., Eds.; Soil Science Society of America: Madison, WI, 1993; Chapter 1, pp 1-26 (104) Fry, V A.; Istok, J D Water Resour Res 1994, 30, 2413-2422 (105) Scow, K M.; Hutson, J Soil Sci Soc Am J 1992, 56, 119-127... Environmental Protection Agency: Washington, DC, April 1993; pp 18-19 (117) Bolick, J J., Jr ; Wilson, D J Sep Sci Technol 1994, 29, 701725 (118) Whiffin, R B.; Bahr, J M In Proceedings of the Fourth National Symposium on Aquifer Restoration and Ground Water Monitoring; National Water Well Association: Worthington, OH, 1985; pp 75-81 (119) Wilson, D J.; Rodrıguez-Maroto, J M.; Gomez-Lahoz, C Sep ´ ´ Sci... 777-784 (98) Fu, M H.; Mayton, H.; Alexander, M Environ Toxicol Chem 1994, 13, 749-753 (99) Robinson, K G.; Farmer, W S.; Novak, J T Water Res 1990, 24, 345-350 (100) Guerin, W F.; Boyd, S A Appl Environ Microbiol 1992, 58, 1142-1152 (101) Summary Report High-Priority Research on Bioremediation; Bioremediation Research Needs Workshop, Washington, DC, Apr 15-16, 1991; U.S EPA: Washington, DC, 1991 (102) Rijnaarts,... R Environ Toxicol Chem 1992, 11, 1197 (109) Landrum, P F Environ Sci Technol 1989, 23, 588 (110) Umbreit, T H.; Hesse, E J.; Gallo, M A Science 1986, 232, 497 (111) Varanasi, U.; Richnert, W L.; Stein, J W.; Brown, D W.; Sanborn, H R Environ Sci Technol 1985, 19, 836 (112) Miller, D D.; Schricker, B R In Nutritional Bioavailability of Iron; Kies C., Ed.; American Chemical Society: Washington, DC, 1981... Rodrıguez-Maroto, J M.; Wilson, D J.; Gomez-Lahoz, C.; Clarke, ´ ´ A N Sep Sci Technol 1995, 30, 521-547 (121) Theng, B K G Formation and Properties of Clay-Polymer Complexes; Elsevier: Amsterdam, The Netherlands, 1979; Chapter 7, pp 157-226 (122) Lightly, J S.; Silcox, G D.; Pershing, D W.; Cundy, V A.; Linz, D G Environ Sci Technol 1990, 24 , 750-757 (123) DeCicco, S G.; Troxler, W L In Standard Handbook of. .. M., Ed.; McGraw-Hill: New York, 1989 (124) Bucala, V.; Saito, H.; Howard, J B.; Peters, W A Environ Sci ´ Technol 1994, 28, 1801-1807 (125) Rodrıguez-Maroto, J M.; Gomez-Lahoz, C.; Wilson, D J.; Clarke, ´ ´ A N Sep Sci Technol 1995, 30, 317-336 (126) Laha, S.; Luthy, R G Environ Sci Technol 1991, 25, 19201929 (127) McCall, P J.; Agin, G L Environ Toxicol Chem 1985, 4, 37 (128) Ma, L.; Selim, H M Water... V L Environ Sci Technol 1988, 22, 377-381 (91) Barriuso, E.; Laird, D A.; Koskinen, W C.; Dowdy, R H Soil Sci Soc Am J 1994, 58, 1632-1638 (92) Boyd, S A.; Jaynes, W F In Layer Charge Characteristics of 2:1 Silicates Clay Minerals; Mermut A R., Ed.; The Clay Mineral Society: Boulder, CO, 1993; pp 48-77 (93) Siantar, D P.; Feinberg, B A.; Fripiat, J J Clays Clay Miner 1994, 42, 187-196 (94) Miller, . use of equilibrium expressions for sorption to natural particles in fate and transport models is often invalid due to slow kinetics. This paper reviews recent research into the causes of slow sorption. Critical Review Mechanisms of Slow Sorption of Organic Chemicals to Natural Particles J O S E P H J . P I G N A T E L L O * A N D B A O S H A N X I N G Department of Soil and Water, The. contain terms for sorption; therefore, an understanding of the dynamics of sorption is crucial to their success. Ignoring slow kinetics can lead to an underestimation of the true extent of sorption,