Environ. Sci. Technol. 2007, 41, 3977-3983
A Binary Aqueous Component Model for the Sediment-Water Partitioning of Trace Metals in Natural Waters ANDREW TURNER* School of Earth, Ocean and Environmental Sciences, University of Plymouth, Plymouth PL4 8AA, U.K.
A model defining the overall sediment-water partitioning of a chemical, KD, and the partitioning of its conservative components, (KD)i, is presented. With respect to many trace metals in natural waters it is proposed that, through strong and perhaps specific complexation, two independent aqueous components coexist and a binary form of the model is appropriate. For two components of a metal that exhibit unequal partitioning, an inverse relationship between KD and particle concentration is predicted. Published experimental measurements of KD for metals in river waters, derived under conditions which exclude variable concentrations of preexistent colloidal particles, displayed either an inverse dependence (Cu, Ni, and Pd) or little dependence (Cs) on particle concentration. Regarding the former, iterative fits with the binary model were better than empirical fits based on a third (colloidal) phase model, and suggested the presence of between about 10 and 75% of a particle-reactive component ((KD)1 ∼ 5 × 104 to 1010 mL g-1) and 25 and 90% of a less reactive (e.g., strongly complexed) component ((KD)2 e 2.5 × 103 mL g-1). Regarding Cs, data indicated the presence of a single component whose KD was on the order of 103 mL g-1. These observations challenge the conventional means by which sediment-water partitioning is considered and modeled, and imply that a third phase is not always a prerequisite for the particle concentration effect frequently observed in laboratory and field studies.
Introduction The partitioning of a chemical between the particulate and aqueous phases is a critical factor when considering its transport and impacts in the environment (1, 2). Conventionally, an empirical distribution coefficient, KD (mL g-1), is employed to define the partitioning:
KD )
Cs CwSPM
(1)
where Cs and Cw represent, respectively, (ad)sorbed and aqueous chemical concentrations on a mass to volume basis, and SPM is the concentration of suspended particulate matter. While a convenient, albeit conditional parametrization (3, 4), in this form the equation considers only total or analytical concentrations of a chemical without discrimination of the different aqueous species and their inherent * Corresponding author phone: +44 1752 233041; fax: +44 1752 233035; e-mail:
[email protected]. 10.1021/es0620336 CCC: $37.00 Published on Web 04/25/2007
2007 American Chemical Society
reactivities toward the particle surface. With respect to many trace metals in natural waters, a multitude of inorganic species and poorly defined organic forms exist. In general, the free ion is considered the most reactive species of metal, and in equilibrium speciation codes, such as WHAM-SCAMP (5), adsorption of divalent metals onto particles is restricted to the free ion, Me2+, and its first hydrolysis product, MeOH+. However, in relatively simple ternary systems, depending on the metal, sorbent, ligand, and pH, organic anions may either stabilize the metal in solution as a relatively soluble complex, or enhance adsorption through the formation of a ternary complex (6, 7). Given these observations, it is reasonable to define an overall equilibrium distribution coefficient as follows:
KD ) RMe2+ (KD)Me2+ +
∑(R
MeX)i((KD)MeX)i
(2)
where RMe2 + and (RMeX)i represent the fractional, equilibrium concentrations of the free ion and remaining, individual aqueous species, respectively, and (KD)Me2+ and ((KD)MeX)i represent the equilibrium distribution coefficients of these forms. Many field and experimental observations of trace metals in natural waters are not, however, fully consistent with the competitive effects and equilibria described above. Regarding aqueous speciation, many metals appear to bind strongly, and sometimes specifically, to certain organic ligands and colloids (including components of the humic/fulvic phase and substances of anthropogenic and biogenic origin); these complexes are, effectively, kinetically inert (8-11). Likewise, metal-particle interactions are often only partly reversible (12) and/or desorption is kinetically constrained (13). More generally, the overall particle-water distribution coefficient exhibits an inverse dependence on sorbent concentration (14-17), an effect that has been ascribed to a variety of processes and artifacts (18, 19). The slow kinetics of metal complexation and, in particular, dissociation have important implications for metal analysis and metal uptake by biota that are becoming increasingly recognized (20-22). However, while the consequences of aqueous phase disequilibria on the adsorption of metals to natural solids have been observed, they do not appear to have been quantitatively or systematically addressed. For example, in several field studies it has been noted that adsorption of many complexing metals is not accompanied by reestablishment of the original aqueous speciation (23, 24), at least within the time scale of experiments or measurements (on the order of a few hours to several days). Moreover, in controlled laboratory experiments, the extent of removal of strongly complexing metals by natural sorbents is dramatically reduced following a period of preequilibration of metal in the aqueous phase (15, 25); thus, different apparent solid-aqueous phase equilibria arise that depend on the sequence of reactant introduction. If it is assumed that strong and/or specific complexes are stabilized in solution (26, 27), such observations suggest that, with respect to adsorption or partitioning, a particle-reactive component of metal and a less reactive, strongly complexed component are, by virtue of thermodynamics or kinetics, largely independent of each other. In a recent paper (28), a model for the octanol-water partitioning of aqueous trace metals in natural waters was developed, based on the assumption that two independent components (groups of species) exist; namely, a hydrophobic component that partitions between the solvent and water and a hydrophilic component that remains in the aqueous VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3977
phase. Excellent fits to experimental measurements of Cu and Pb in a number of river-water samples were obtained using this model, reinforcing the validity of the underlying assumption. A similar approach has been postulated by van Leewen and Pinheiro (22) in which the free ion and lipophilic complexes are considered independently during metal biouptake. To this end, the current paper explores the possibility and likely consequences of pooling trace metals to model their net partitioning between suspended sediment particles and water in natural environments. Thus, it is hypothesized that there coexist a particle-reactive component of metal, containing the free ion and relatively reactive complexes, and a second, independent component, including strong and perhaps specific complexes, that has relatively little propensity to interact with the particle surface. A mass balance model based on this concept is developed below, and its applicability is examined using empirical data for a number of trace metals in a variety of river-water environments reported in the literature.
Theory The Binary Aqueous Component Model. The model is based on the relationship between the overall distribution coefficient of a number of conservative, noninteracting chemicals and the individual distribution coefficients of its components. The underlying principle is, therefore, analogous to that inherent in other multicomponent models (22, 28, 29). Using current notation, and with respect to a given trace metal, the total concentration of each individual species, i, is as follows:
Ci ) (Cw)i + (Cs)i
(3)
and the distribution coefficient of each species is thus:
(KD)i )
(Cs)i
(4)
(Cw)iSPM
Aqueous and (ad)sorbed concentrations can be rewritten in terms of KD and SPM as follows:
(Cw)i )
Ci
(5)
1 + (KD)iSPM
(Cs)i ) Ci -
Ci 1 + (KD)iSPM
(6)
Assuming component conservation, the overall distribution coefficient, KD, is the ratio of summed particulate concentrations to summed aqueous concentrations:
(
1-
KD )
Ri
∑ 1 + (K ) SPM D i
Ri
∑ 1 + (K ) SPM D i
)
1 SPM
(7)
where here, total concentration has been replaced by a fractional term, R. Regarding trace metals in natural waters, where specific ligands and/or strong complexes that are slow to dissociate compared with the time scale of adsorption are involved, it is appropriate to consider multiple groups of species that are effectively independent. To a good approximation, two components may be invoked, as conceptualized in Figure 1. Thus, component 1 contains forms that are (relatively) particle-reactive, shown by the free ion, Me2+, and a variety of complexes, MeX. These species are implicitly at equilibrium with each other such that consumption of Me2+ at the particle surface is compensated by relatively rapid dissociation of 3978
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 11, 2007
FIGURE 1. Conceptualization of the binary aqueous component model for trace metal partitioning in natural waters. Component 1 is indicated by the free ion, Me2+, and complexes with X2-, and component 2 is indicated by independent, strong complexes with L2-. MeX, and the net partitioning of this component reflects the weighted average partitioning of all species (an effect defined by eq 2). Component 2 comprises (relatively) unreactive forms, indicated by strong complexes with L2-, that are independent of component 1. For two conservative components, eq 7 may be rewritten in binary mode as follows:
KD )
(
R1
)
R2 1 + (KD)1SPM 1 + (KD)2SPM 1 (8) R1 SPM R2 + 1 + (KD)1SPM 1 + (KD)2SPM
1-
Provided that (KD)1 > (KD)2, a consequence of this model is an inverse dependence of the overall distribution coefficient on SPM concentration, an effect that is commonly observed for trace metals (and other chemical constituents) in natural waters (14-19). This arises because, with increasing SPM concentration, the reduction in the absolute aqueous concentration of component 1 required to maintain equilibrium with the particulate phase is accompanied (and offset) by a smaller reduction in the aqueous concentration of component 2. Figure 2a illustrates the particle concentration effect for hypothetical, equal mixtures of component 1, whose net distribution coefficient, (KD)1, is 105 mL g-1, and component 2, whose net distribution coefficient, (KD)2, is varied between 102 mL g-1 and 104 mL g-1, over a particle concentration range typical of rivers and estuaries. The effect is illustrated from a different perspective in Figure 2b. Here, KD is shown as a function of particle concentration for different relative proportions of the two components ((KD)1 ) 105 mL g-1; (KD)2 ) 102 mL g-1). Note that increasing (KD)1 is accompanied by a reduction in the magnitude of the term: R1/(1 + (KD)1SPM); such that, for a highly reactive first component, the relationship may be approximated by an equation of the form: (KD)1 ) m/SPM + c; where m and c represent the gradient and horizontal asymptote ()(KD)2/R2), respectively. According to the binary aqueous component model, and with respect to trace metals, the magnitude of the particle concentration effect is predicted to increase on increasing (i) the difference between the reactivities of the two components ((KD)1/(KD)2), and (ii) the proportion of the strongly complexed component (R2). In practice, eq 8 can be solved by iteration for the relative abundance and reactivity of both components from measurements of the overall distribution coefficient as a function of particle concentration. The Third-Phase Model. The binary model derived above bears some semblance to the “third-phase” model (17, 29, 30), which is commonly invoked to explain the particle
FIGURE 3. Overall distribution coefficient versus particle concentration for hypothetical mixtures of particles and a third phase, calculated using eq 9 and where (KD*)1 ) 105 mL g-1, (KD*)2 ) 106 mL g-1, and r* is varied as annotated.
FIGURE 2. Overall distribution coefficient versus particle concentration calculated using eq 8 for (a) an equal mixture of two independent components of the same metal, where (KD)1 ) 105 mL g-1 and (KD)2 is varied as annotated, and (b) mixtures of two independent components of metal, where (KD)1 ) 105 mL g-1, (KD)2 ) 102 mL g-1, and r2 is varied as shown. concentration effect. Here, two aqueous pools of metal (truly dissolved and colloidally bound) are discriminated on a size basis, and an inverse dependence of KD on SPM arises because it is assumed that the concentration of colloidal particles, operationally encompassed by the aqueous phase but acting as an additional sorbent or complexant, is a function of SPM concentration. The mathematical formulation of the thirdphase model is as follows (17):
KD )
(KD*)1 1 + (KD*)2R*SPM
(9)
where here, (KD*)1 and (KD*)2 are the distribution coefficients defining the partitioning of the chemical between sediment and water and the colloidal phase and water, respectively, and R* is the fraction of the third phase relative to the concentration of SPM. Accordingly, R*SPM is the w/v concentration of the third phase, although it should be mentioned that proponents of the model sometimes infer a nonlinear dependence of its concentration on SPM (17). Unlike the binary aqueous component model, which considers the partitioning of two independent components of metal with respect to a single solid phase, the distribution coefficients defined in the third-phase model refer to the equilibrium partitioning of total metal with respect to two independent solid phases, one of which is encompassed by the aqueous phase. It is important to note that the binary aqueous component model does not exclude the possibility of colloidal particles, but requires that they are invariant in concentration and contribute, as complexants or even sorbents (11), to the aqueous fraction of component 2. In Figure 3, KD is plotted against SPM concentration using respective values of (KD*)1 and (KD*)2 of 105 and 106 mL g-1, and for different relative concentrations of the third phase.
Note that, unlike the binary model, the horizontal asymptote of the relationship is zero. Note also that over a broad range of conditions the relationship is approximately linear on logarithmic axes. Consequently, the third phase model is often empirically defined as: logKD ) -blogSPM + loga; where a and b are constants (14, 19). By depressing R* or the distribution coefficient of the third phase, the magnitude of the particle concentration effect is reduced and the nonlinear portion of the relationship is shifted to higher particle concentrations. (A logarithmic relationship is approximated in the binary aqueous component model under conditions dictated by the absolute and relative magnitudes of the controlling variables, and as exemplified in Figure 2.)
Data Derivation The data examined in this study have been generated as part of previous research programs undertaken in our laboratory or have been taken from the literature and relate to the sediment-water partitioning of the strongly complexing trace metals, Cu, Ni, and Pd, and, for comparison, the weakly complexing Group IA metal, Cs, in a variety of river waters (characterized in Table 1). Data are not exhaustive, but are specific to a broadly consistent experimental approach that minimizes artifacts associated with variations in particle character and concentrations of preexistent colloidal (filterable) particles. With respect to our experiments examining the partitioning of Ni and Pd, Beaulieu or Tamar river water and respective samples of oxic bed sediment were collected concurrently in individual polyethylene bottles. Water was vacuum-filtered through 0.45 µm cellulose acetate membrane filters, while sediment was washed and sieved through a 63 µm Nylon sieve using filtered river water. The fine fraction of sediment was then retained by 0.45 µm filtration to remove excess water and any confounding aqueous reactants. These include colloids and other potential complexants that preexisted in the interstitial waters or that were formed (or desorbed) during sieving in indigenous water. Particles retained by (but not occluded in) the filter, now between 0.45 and 63 µm in diameter, hence, in theory, free of colloids, were resuspended in 50 mL of filtered river water as a slurry whose SPM concentration was about 50 g L-1. Meanwhile, the analyte was added, individually, as either a radiotracer (63Ni) at a concentration of few nM, or as stable metal (Pd) at a concentration of a few tens of nM, to identical aliquots of filtered river water (either 50 or 100 mL in polymer bottles), and the contents were allowed to equilibrate for a period of up to 24 h. Different quantities of slurry were then added to attain a range of SPM concentrations (up to about 1 or 2 g L-1) of a single, albeit heterogeneous, particle population. In VOL. 41, NO. 11, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3979
TABLE 1. Chemical Characteristics of the Water Samples and Means of Particle Isolation Adopted in the Partition Studies metal, river
pH
Ni, Tamar Pd, Tamar Ni, Beaulieu Cu, Susquehanna Cs, Tamar
7.70 7.93 7.25 7.46 7.20
Cs, Meused
DOC, µM SPM, mg L-1a 200 320 750 470 310 0
