Environ. Sci. Technol. 1993, 27, 1388-1393
Kinetic Study of the Speciation of Copper(I1) Bound to a Hydrous Ferric Oxide Donald W. Gutzman and Cooper H. Langford’
Department of Chemistry and Biochemistry, Concordia University, 1455 d e Maisonneuve Boulevard West. Montreal, Quebec, Canada H3G 1M8 The equilibrium distribution of metal ions over distinguishable sites of a heterogeneous complexant may be determined by monitoring the rate of reaction of the equilibrated system with a ligand capable of stripping the metal ion off the various sites. This kinetic speciation method has the advantage that it indicates not only the equilibrium distribution but also the relative lability of complexes. It may also reveal lability differences among complexes of very similar thermodynamic stability. This kinetic method has been applied to the speciation of Cu(11)in the presence of a hydrous ferric oxide colloid using a copper-selective color-forming chelating ligand. The pH (4.5-7.0)range spans the region over which an “adsorption edge”is observed in conventional isotherm measurements. The kinetic speciation results are consistent with constraints suggested by these isotherm experiments. A main point of this study is that significant differences in speciation are observed for solutions where Cu(I1) was added to already-formed colloids in contrast to those in which Cu(I1) was present during the formation of colloids. Differences between equilibrium and kinetic speciation results may be explained by hypothesizing differential diffusional accessibility to similar binding sites. Introduction In aqueous systems, Cu(I1) is an essential nutrient at appropriate concentrations and becomes a toxicant at elevated levels (1-3). From a toxicological point of view, risk is a function of both toxicity and bioavailability (47).A recent proposal is that toxicity may be more closely related to the concentration of the “free ion” ([Cu(H20)e2+])than the total metal concentration. This is the so-called ”free ion-hypothesis” (8). It may be that this hypothesis relates closely to the rate of release of the metal ion from complexes as much as to the equilibrium speciation since rate of uptake by the organism is critical, but little work has been done on this point. [The work that has been published focuses on convenient “laboratory ligands” for Cu(I1) which do not occur in nature.] In a number of natural water cases, equilibration times are long compared to rates of uptake by organisms so that the organism samples a nonequilibrium distribution of species. This paper describes a part of a program which makes an approach to the goal of kinetic speciation. Progress to date has been reviewed (9). The major ligands binding trace metals in natural waters are hydrous oxide colloids and humic substances (colloidal organic acids). We report elsewhere on Cu(I1) speciation in complexes of a wellcharacterized humic acid (10). Here the study concerns copper binding to hydrous ferric oxides. The results of
* To whom correspondence should be addressed at his present
kinetic speciation suggest differentiation which is not observed at equilibrium. This may arise for thermodynamically similar sites of different diffusional accessibility. Theory of the Method Consider the general case of a metal, Mz+,associated with a heterogeneous ligand system with binding sites L,. The equilibrium mixture is treated with a strongly complexing chromophoric ligand, R, which can strip the metal from Ln sites and form a detectable complex. The reaction of R with kinetically distinguishable Ln-M complexes leads to a common product, MR, as shown in the following competing reactions. ko
M ~ ++ R-MR Ll-M
+R
L2-M
+R
L,-M
iR
*
ki +
kz +
kn +
(1.1)
MR + L1
(1.2)
MR + L2
(1.3)
MR + L,
(1.4)
Use of a large excess of R reduces these reactions to pseudofirst-order kinetics. The kinetically important parameters of pH and ionic strength are buffered by the choice of the composition of the solution containing the chromophore, R. Formation of the common product may then be described by the equation:
where C(t) is the concentration of MR at time t, C O is , ~the concentration of the ith component of the mixture at time = 0, and ki is the rate constant for reaction of the ith component with R, which is at least pseudo-first-order and may be intrinsically first order if a disjunctive mechanism (11)applies. X is a time-independent signal which combines background or blank absorbance with absorbance due to the reaction of components which react too rapidly for the experiment to measure. As eq 2 implies, nonlinear regression of the C(t) to yield initial species concentrations and rate constants is required. But first, an appropriate number of distinguishable kinetic components must be identified. We adopt an approachdevised by Shuman and co-workers (12,13).This method uses a numerical approximation of the Laplace transform: C(t) = SJH(k)e-kt d(ln k)
address: Department of Chemistry,The University of Calgary,2500 University Dr. N.W., Calgary, Alberta, Canada T2N 1N4.
where H ( k ) d(ln k) represents the probability of an ML site reacting with a rate constant between In k and In k
1388 Envlron. Scl. Technol., Vol. 27, No. 7, 1993
0013-936X/93/0927-1388$04.00/0
0 1993 Amerlcan Chemical Society
+
d(ln 12). The second-order numerical approximation takes the form:
H(12)= d2C(t)/d(lnt)’ - dC(t)/d(ln t)
Plotting H ( k ) vs In t results in a kinetic spectrum. Peaks in the spectrum indicate the number of distinguishable components and provide estimates of the associated rate constants. The areas under the peaks are proportional to concentrations. Unfortunately, the second-order numerical Laplace transform is sensitive to artifacts. The best procedure is to use the parameters derived, judiciously, as input to the nonlinear regression. The chemical significance of the results is then accepted only if they pass tests of chemical consistency as pH and concentration ratios in equilibrated samples vary. The usefulness of this approach has been demonstrated in studies of Ni(I1) bound to a fulvic acid (14) and of Cu(I1) bound to humic acid (10). A particularly simple mechanism often underlies eq 1. It is called “disjunctive” (11) kd
L,M -,M”’ fast
+ L,
M ~+ + R -, MR In this case, the apparent first-order rate constants obtained in the presence of excess R are simply the rate constants for dissociation of the species under study at the pH and ionic strength of the kinetic experiment. The disjunctive mechanism is expected to apply for simple monodentate and bidentate complexing sites. Ita alternate, the “adjunctive” mechanism is characteristic for multidentate chelating ligands. There is good reason to believe that the sites for binding Cu(I1) to hydrous ferric oxides and humic substances are mono and bidentate (15, 16). Experimental Section
Reagents. All glassware used in reagent preparation was soaked for 24 h in 10% nitric acid followed by rinsing with doubly distilled water to minimize metal ion contamination. A higher grade water from a Sybron/Barnsted NANOpure-A system was used for a final rinse and all solution preparation. Unless otherwise specified, all chemicals were of reagent grade and were used without further purification. Stock HC1 (1.000 M) was prepared from Acculute (Anachemia)vials. Solutions of approximately 1.0 X M CuSO4 in 1.00 X M HC1were prepared using Fisher CuS046H20 which had been recrystallized from pure water. Iron(II1)solutions [approximately 1.0 or 2.0 x were prepared by dissolving FeC13 (Fisher) in 1.00 X M HC1. Stock solutions of the weak base “Tris” [tris[hydroxymethyl)aminomethanelwere prepared at 0.500 M from Sigma chemicals “trizma base”. Colloids. Hydrous iron oxide colloids were prepared by the continuous dropwise addition of dilute Tris to stirred 50.0-mL mixtures of Fe(II1) and Cu(I1) solutions
in 1.00 X 10-2 M HC1. A peristaltic pump was used to deliver Tris to four flasks at once. All four solutions were prepared at a single Fe(II1) concentration while Cu(I1) loading was varied. All sampleswere at room temperature and open to the air. Addition continued for 4-6 h until the total volume was about 98 mL. pH was adjusted manually to precise desired values by the addition of further Tris. The volume was then adjusted to 100.0 mL with water and the pH verified. One blank Fe-containing colloid with no Cu(I1) loading was prepared each time. For some colloids, Cu(I1) was added after the hydrous ferric oxide was formed. These are designated “postaddition” colloids below. These were prepared by addition of suitable quantities of the Cu(I1) stock solution to 25.0mL samples of dispersions of the “blank”colloids described above. Colloids were prepared with total iron concentrations of approximately 1,2, or 3 x lo4 M at pH values of 4.5, 5.5, 6.0, 6.7, and 7.0. Mole ratios of Fe:Cu ranged from about 2 to 30 with the higher values only for higher total iron concentrations. That is, Cu(I1) concentrations were required to be maintained at levels high enough to permit good resolution of absorbance changes in kinetic analysis. It is to be noted that all Cu(I1) concentrations are well below solubility limits for solid Cu(I1) phases at these pH values. A “standard” with an Fe:Cu ratio of 6.5 was prepared with each set of four samples to monitor the consistency of preparation. All samples were aged sealed in the dark at room temperature for approximately 24 h prior to kinetic analysis. Homodisperse hematite colloids (a-Fe203) were the generous gift of Prof. L. K. Koopal, Wageningen Agricultural University, The Netherlands (17). Binding of Cu(I1) to the 212-nm diameter particles was studied by adjusting the pH of a dispersion and Cu(I1) to pH = 6.00 with Tris. The sample was left to equilibrate for 24 h before kinetic analysis. Kinetics. Reaction kinetics were followed using a Hewlett-Packard 8452A diode array spectrometer in a single wavelength absorbance-time mode. Absorbance was typically measured at intervals of 3 s with an integration period of 1 s. Mixing was performed by the addition 1.00 mL of chromophore/buffer solution to 1.00 mL of colloid sample in a 1.00 cm path cell at time zero. Solutions were further mixed with a Pasteur pipet. Both solutions were preequilibrated to the thermostat temperature of 25.0 f 0.01 “C which was maintained in the cell chamber with a circulating water bath. The chromophoric ligand used for the kinetic speciation of Cu(I1) was 3-propyl-5-hydroxy-5-(D-arabino-tetrahydrobutyl)thiazolidine-2-thione(PHTTT). It was prepared bythemethodofChoweta1. (18). Atthe pHofthepresent kinetic studies, the ligand has amolar extinction coefficient of 13100 i 300 M-l cm-l at 434 nm. Aqueous chromophore/buffer solution was prepared using Bis-Tris [ [bis(2-hydroxyethyl)aminoltris(hydroxymethy1)methane;Sigma Chemical], HC1, PHTTT. A combination of equal volumes of the chromophore/buffer solution with the Cu(I1)-containing hydrous ferric oxide M PHTTT at sample results in a mixture of 2.50 X an ionic strength of 0.100 and buffered at pH = 5.7 (25.0 “C). This pH was chosen because it lies at the center of the range of the samples studied (4.5-7.0). The reaction of free Cu(I1) was much too fast to measure under these conditions, as expected. The Cu-PHTTT complex was Envlron. Scl. Technol., Vol. 27, No. 7, 1993 1389
stable over periods longer than the 60-120-min duration of the kinetic analyses. Results and Discussion
The hydrous ferric oxide used in this work was prepared by the slow hydrolysis of hexaaquairon(II1). The formation and properties of hydrous ferric oxide colloids have received considerable attention (for example, see refs 1923). The slow hydrolysis approach is simple to execute and control in the laboratory and is, therefore, popular. It is unlikely, however, that the majority of colloids found in natural water systems arise this way. It is more probable that the majority arise from the slow oxidation of Fe(1I) as it diffuses up from the anoxic zone. Unfortunately, laboratory preparation of consistent colloids by this oxidative method is difficult. Equilibrium Method of Speciation of Cu(I1). Some of the hydrous ferric oxides prepared in the presence of Cu(I1) were analyzed for free Cu(I1) using a separation method. This has been previously reported (24). After the 24-h aging period, a suitably small fraction of the sample to avoid perturbation of the species distribution was collected through a micropore filter. The pore size (0.45 pm) was small enough to prevent the passage of measurable Fe as confirmed by atomic absorption analysis. The copper concentration of the filtrate was taken to indicate the free Cu(I1)concentration not bound to colloids. This equilibrium data was fitted quite satisfactorily to a Langmuir isotherm which models with only one independent binding site type and a single binding constant. Thus, these equilibrium measurements were not able to distinguish among different binding sites for Cu(I1) on the colloid. The experiments offer no concrete hint of site heterogeneity. Kinetic Speciation of Cu(I1). The various samples examined had Fe and Cu total concentrations varied over a factor of 5. The range of Fe:Cu ratio was varied from 1.6 to 26. pH was varied from 4.5 to 7.0, covering the region of the “adsorption edge” of hydrous ferric oxides. Under each condition, the equilibrated sample was reacted with the PHTTT/buffer solution so that the kinetics of release of Cu(I1) from each equilibrated sample was recorded under constant kinetic conditions by conventional spectrophotometry. A typical absorbance-time curve from these experiments is shown in Figure 1. A smoothed version of this curve suitable for differentiation was obtained by least-squares fitting of a polynomial. It is shown in Figure 2 in log form with the raw data superimposed as the noise on the smoothed curve. All smoothed runs were analyzed using the Laplace transform method to identify the number of components and the preliminary trial values of rate constants and concentrations (13,141. Figure 3 shows the Laplace transform of the data from Figures 1 and 2. These estimates were then used in the nonlinear regression routine to refine the values. Two criteria are imposed before we present values obtained as having any chemical significance (since there is always a danger that real data with noise are simply successfully archived by a set of fitting constants not described by genuine chemical parameters). The first of these is that the rate constants obtained must be stable as concentration ratios and pH vary, indicating a consistent set of chemical types in the. The second criteria is that the concentrations of these components must vary in a way consistent with the law of mass action considerations. 13gO Environ. Scl. Technol., VoI. 27, No. 7, 1993
0
1000
8000
4000
SO00
Time (seconds) Flgure 1. Conventionalabsorbancemonltorlngat 434 nmof the reaction between Cu(II)-containing hydrous ferric oxide (Fe:Cu mole ratio = 19.3)prepared at pH 5.5 and the chromophore PHTTT. Although the kinetic change in absorbance is fairly small, the noise level is low In this case.
. . . . I .. . . I . . . .
-8.00
,. . . q-n. . ,.. . .,. . . . #
-2.80
E
3
4
6
6
7
8
9
In(time) Flgure 2. Smoothing applied to the absorbance-time data shown in Figure 1. Data has been fit as in (A434) vs In (r) using a 9th degree polynomial. This is the lowest degree polynomialwhich faithfully folbwed the contours of the curve.
Stability of Rate Constants. The results shown in Table I illustrate the degree of stability of the rate constants. Two components with reactions on the time scale slow enough for conventional mixing kinetics were consistently identified. The variations over the full concentration ratio and pH range are small enough that they cannot be said to differ in a statistically significant way. Both rate constants appear to be “rugged” with average values of 0.028 f 0.010 and 0.0014 f 0.0006 a-l. The apparent small trend in k3 values with pH may be a consequence of the better resolution of the data at low pH. Generally, noise levels were observed to decline with declining equilibration pH. In a few runs having particularly low noise, the Laplace transform hinted at the presence of three resolvable components in the accessible time domain. If this is in fact true, it would account for the variation observed in the two rate constants used in the nonlinear fits. The example in Figure 3 is one of the cases where the intermediate peak is seen. This underlines the view we take of the rate constants reported. They are not expected to correspond to two precisely defined and homogeneoue
k=0.0014 C=0.0097
C-0.0094
0
"X" Componen
D Component 2
v Component 3
0.001
A
R
Not recovered
SO * O F
0.003
0.00%
Fe:Cu m o l e ratio
In(time) Figure 3. Laplace spectrum of the data shown In Flgures 1 and 2. Normally Laplace analysis Indicated the presence of only two klnetlc components. When absorbance-time data were less nolsy, a third component could sometimes be found. It appears that the k2 normally measured may actually be a combinatlon of two poorly resolved components.
Table I. Rate Constants for Species of Intermediate Lability Related to Kinetic Speciation of Copper(I1) in Iron(II1) Hydrous Oxide system Cu-Fe hox: preaddition
colloid pH n1 n2
rate constants (9-9
ks
kz
4.5 1 5 0.024 i 0.007 5.5 3 11 0.032i 0.010 6.0 4 13 0.028 i 0.008 6.7 3 14 0.024 0.011 7.0 1 4 0.032i 0.016 4.5-7.0 12 47 0.028f 0.010 postaddition 4.5-6.7 7 9 0.032 0.008 Cu-hematite 6.0 1 2 0.0478i0.0005
0.0012 f 0.0002 0.0014i 0.0002 0.0013 0.0002 0.0016 O.OOO9 0.0016 t 0.0005 0.0014 f O.OOO6 0.0018i 0.0010 0.0018+0.0005
**
a Results have been determined from data measured by conventional spectrophotometry. n1 = number of colloid preparations. n2 = number of successful Laplace/NLR results.
site types. Rather, they probably represent averages over a distribution of similar sites which remain kinetically defined over the full range of conditions. The peaks in the numerical Laplace transform are not sharp for two well-defined Components as a result of numerical approximations in the equations. It is, therefore, difficult to distinguish a distribution of similar sites from a few welldefined sites in the presence of noise. However, we believe that the Laplace peaks are broad enough to support the distribution interpretation. Concentration data also support the presence of a minor third component class. When only two components are used in the analysis,the s u m over all fitted concentrations is somewhat less than the raw estimate of the change in concentration associated with the absorbance change from If a third a manually extrapolated t = 0 to t = component is used, the agreement with this approximate parameter is significantlybetter. This simply underlines the key point, the number of components classes report is the minimum number giving a statistically adequate representation of the system. One of the prime aims of our program is to conduct speciation studies a t concentrationswhich approach values realistic for natural waters. The total absorbance change QD.
Flgure4. Component contrlbutbnsfor the klnetkspedetion of copper(11) from hydrousferrlc oxlde at pH 4.5 determined ?rOm Ineesurement of the absorbing complex Cu(PHmh by conventlonal spectrometry. The majority of copper is In the hlghly labile "X' component even at larger Fe:Cu mole ratios.
associated with kinetically observable components was as little as 0.01 absorbance units. The relative noise levels are considerable and pose a formidabletalkfor the Laplace/ NLR analysis. The choice of higher total concentrations might have led to the regular resolution of a third component, but at the price of moving away from environmental relevance. As we have seen in experiments (to be reported elsewhere) using laser thermal lensing to reduce the required total Cu(II), speciation is dependent upon total concentration. Absorbance measurements on solutions prepared by postaddition of Cu(I1) were generally very noisy. Very few runs allowed clean Laplace/NLR analysis. Those which were analyzable yielded values for the critical rate constants k2 and k3 similar to those from the case where Cu(II) was present during colloid formation (Table I).The similar Cu(1I) labilities suggest that Cu(I1) can diffuse into the particles, but we see below that the distribution of concentrations of components is significantly different in the pre- and postaddition cases. That is, the sites appear to be the same, but distribution of Cu(I1) over themvaries according to the history of the colloid. Table I also presents some rate constants determined from studies of Cu(I1)added to the homodisperse hematite particles of 212-nm dimension. These were studied after a 24-h equilibration at pH = 6.0. The slower of the two rate constante obtained is quite similar to ks above. The faster rate constant is somewhat larger than k2 but is similar to the largest rate constant determined under conditions where three components were resolved. Some caution is warranted in drawing conclusions, the signals were quite noisy in the first 500-1000 s. Chemical Significanceof Components. The master variable used in studying speciation of Cu(I1) bound to hydrous ferric oxides has been pH. Figures 4 and 5 show the results for the samples equilibrated at pH of 4.6 and 6.7-on both ends of the well-knownadsorption edge. Error bars are shown for a few illustrative points (to maintain overall readability of the figures) and indicate standard deviation in five determinations using different total Fe and Cu concentrations and different colloid preparation batches. Error bars are typical with the exception of pH 5.5 experiments where they may be somewhat larger (uide infra). Environ. Scl. Technol., Vol. 27, No. 7, 1993 13Sl
...," " , " " " . I ' " . I ' ' ' "X" Componen
DH
6.7
0
Component 2
This case is the parallel to the classic absorption edge experiments (25,26). This is not really inconsistent with the report of Swallow et al. (27),who observed that only slightly less Cu is carried down in a precipitation process if the Cu was added soon after precipitation than if it was present during the formation of the precipitate. Our results suggest that the binding is different in character.
Conclusion
Fe:Cu m o l e ratio Figure 5. Species distribution of Cu(I1) in hydrous ferric oxide equilibrated at pH 6.7. Note that on this high pH skle of the adsorption
edge much less copper is present in a highly labile form and significantly more is not recoverable on the time scale of the experiments.
Notice that these figures include as the major contributor the component which is the time-independent absorbance due to Cu(II), which is bound to PHTTT within the time of mixing. Free Cu(H20)s2+will be the major contributor, but some simple mono- and bidentate complexes at the surface are also expected to be sufficiently labile to appear in this term. This concentration is somewhat larger than free C U ( H ~ O )found ~ ~ + by separation (24). The components associated with the rate constants k2 and kS are probably sites located within the particles. A single line is used to represent these components in the figures as their contributions were quite similar. The figures also show that a fraction of the Cu(I1) was insufficiently labile to be recovered within reasonable reaction time. The trend of species distribution with Fe:Cu ratio is similar at each pH. As Fe:Cu increases, the proportion of free or quite labile Cu(1I) ("X"component)decreaseswhile the amount of Cu bound at less labile, and presumably thermodynamically more stable sites, increases. The largest single contributor to the less labile Cu(I1) is that which is incorporated into the hydrous oxide lattice sufficiently strongly so as not to be recoverable on the time scale of Cu(I1) reactions with PHTTT, free of Fe interference. Literature reports of studies of Cu(I1) adsorption onto an excess of preformed, macroscopically manipulatable, hydrous ferric oxide particles (25, 26) show copper adsorption changing from 10% to 90% adsorbed over the adsorption edge centered at pH = 5.5 with a width of about 1pH unit. Our results indicate a change from about 20 % bound to about 70% bound across the edge pH range. The edge pH does depend on the Cu:Fe ratio. This is probably the main cause of this difference. The high noise we encountered in pH = 5.5 measurements is probably due to the location of this pH at the most sensitive region of the edge. The comparison of the main body of results to those where copper was added after colloid formationdoes reveal some interesting differences in the distribution of components. In the postaddition case, the labile "X" fraction is a more important component at the expense of the nonrecoverable, which tended to zero. It would appear that Cu(I1) is more likely to be surface bound in this case and unlikely to be integrated into the hydrous oxide lattice. 1392 Environ. Sci. Technol., Vol. 27, No. 7, 1993
The literature of binding Cu(I1) to hydrous oxides is mainly that of studies by equilibrium methods where a single binding constant and a homogeneous set of binding sites were an adequate model. This has been interpreted in terms of a dominant type of binding site at low coverage. Kinetic methods seem to suggest greater heterogeneity. This does not necessarily imply a conflict. If the principle cause of kinetic heterogeneity is not the strength of binding site but the mass transport limitations on the rate at which Cu(I1)becomes available to PHTTT, sites of similar overall thermodynamic stability could be differentiated because both the rate of formation of the complex and ita dissociation may be limited by diffusion to the site. We see that the addition of Cu(I1) before the colloid forms leads to greater loss which is plausibly interpreted in terms of Cu(I1) occupation of sites in the interior of the particles, possibly in the oxide lattice.
Literature Cited (1) Eichenberger,E. In Concepts on Metal Zon Toxicity;Sigel, H., Ed.; Marcel Dekker: New York, 1986;pp 67-100. (2) Sunda, W. G.; Hanson, A. K. Limnol. Oceanogr. 1987,32. (3) Martin, R.B.In Concepts on Metal Zon Toxicity; Sigel, H., Ed.; Marcel Dekker: New York, 1986,pp 21-65. (4) Sunda, W. G.; Guillard,R. R. L. J. Mar. Res. 1976,34,511529. ( 5 ) Sunda, W. G.; Lewis, J. A. M. Limnol. Oceanogr. 1978,23, 870-876. (6) Anderson, D. M.; Morel, F. M. M. Limnol. Oceanogr. 1978, 23,283-295. (7) Petersen, R. Environ. Sci. Technol. 1982,16,443-447. (8) Lund, W. Fresenius J . Anal. Chem. 1990,337,557-564. (9) Langford, C. H.; Gutzman, D. W. Anal. Chim. Acta 1992, 256,183-201. (10) Bonifazi, M. M.Sc. Thesis,ConcordiaUniversity,Montreal, 1991. (11) Hering, J. G.;Morel, F. M. M. Environ. Sci. Technol. 1990, 24, 242-252. (12)Shuman, M. S.;Collins, B. J.; Fitzgerald,P. J.; Olson, D. L. In Aquatic and Terrestrial Humic Materials; Chrietman, R. F., Gjessing, E. T., Eds.; Ann Arbor Science: Ann Arbor, MI, 1983,pp 349-370. (13) Olson, D. L.; Shuman, M. S. Anal. Chem. 1983,55,11031107. (14) Lavigne, J.A,; Langford, C. H.; Mak,M. K. S.Anal. Chem. 1987,59,2616-2620. (15) Schindler, P. W. In Circulation of Metals in the Enoironment; Sigel, H., Ed.; Marcel Dekker: New York, 1984; p 111. (16) Gamble, D. S.;Underdown, A. W.; Langford, C. H. Anal. Chem. 1980,52,1901-1908. (17) Penners, N. H. G. The preparation and stability of homodisperse colloidal haemutite (a-FezOs). Ph.D. Thesis, Wageningen Agricultural University, The Netherlands, 1985. (18) Chow, Y. M.; Mak, K. S.;Lau, 0. W. Analyst 1977,102, 139-140. (19) Langmuir,D.;Whittemore,D. 0.InNonequilibrium systems in natural water chemistry; Advances in Chemistry Series 106;American Chemical Society: Washington, DC, 1971.
(20) Sommer, B. A.;Margerum, D. W.; Renner, J.; Saltman, P.; Spiro, T. G. Bioinorg. Chem. 1973,2,295-309. (21) Dousma, J.; de Bruyn, P. L. J. Colloid Interface Sci. 1976, 56,527-539. (22) Byme, R. H.; Kester, D. R. Mar. Chem. 1976,4,256-274. (23) Buffle, J. Complexation reactions in aquatic systems: An analytical approach; Ellis Horwood Chinchester, 1988; pp 330-338 and references cited therein. (24) Gutzman, D. W.;Langford, C. H. Water Pollut. Res. J . Can. 1988,23,379-387.
(25) Davis, J. A.;Leckie, J. 0.Environ. Sci. Technol. 1978,12, 1309-1315. (26) Benjamin, M. M.;Leckie, J. 0. J. Colloid Interface Sci. i98i, 79,209-221. (27) Swallow,K. C.; Hume, D. N.;Morel,F. M. M.Environ. Sci. Technol. 1980,14,1326-1331.
Received for review October 22, 1992.Revised manuscript received March 16, 1993.Accepted March 24, 1993.
Envlron. Scl. Technol., Vol. 27, No. 7, 1993 1993
