X-ray absorption spectroscopy of Co(II) sorption complexes on quartz (α-SiO2) and rutile (TiO2)

Geochimica et Cosmochimica Acta, Vol. 60, No. 14, pp. 2515-2532, 1996 Copyright © 1996 Elsevier Science Ltd Printed in the USA. All rights reserved 0016-7037/96 $15.00 + .00

Pergamon

P I I S0016-7037(96) 00114-7

X-ray absorption spectroscopy of Co (II) sorption complexes on quartz (~-SiO2) and rutile (TiO2) PEGGY A. O ' D A y , I CATHERINE J. CHISHOLM-BRAUSE, 2 STEVEN N. TOWLE, 3 GEORGE A. PARKS, 3 and GORDON E. BROWN JR. 3'4

t Department of Geology, Arizona State University, Tempe, AZ 85287-1404, USA 2Department of Physical Sciences, Virginia Institute of Marine Science, College of William and Mary, Gloucester Point, VA 23062-1346, USA 3Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA 4Stanford Synchrotron Radiation Laboratory, P.O. Box 4349, MS 69, Stanford, CA 94309-0210, USA (Received July 22, 1995; accepted in revised form March 26, 1996)

Abstract--The local molecular structure of Co(II) surface complexes sorbed to two pure mineral substrates, quartz (a-SiO2) and rutile (TiO2), was examined with X-ray absorption spectroscopy (XAS) and extended X-ray absorption fine structure (EXAFS) analysis. Absorption spectra were collected for samples equilibrated with Co solutions under- and over-saturated with respect to solid Co-hydroxide phases (0.15-3.00 mM) at Co surface coverages of 0.63-9.51 #mol m -2 and for equilibration times of 23 h to 21 days. Quantitative analysis of the EXAFS spectra indicates no significant structural differences in the local atomic environment around Co sorbed to quartz, regardless of the surface coverage or the equilibration time. For all Co/quartz samples, the local Co environment is similar (but not identical) to that of Co in solid, crystalline cobalt hydroxide (Co(OH)2(s)). Cobalt is octahedrally coordinated by six O atoms at 2.06-2.11 A; second-neighbor Co backscatterers are found at 3.11-3.12 ,~, slightly contracted relative to the Co-Co distance in Co (OH)2 (s) ( 3.173 A). These results and the analysis of Co multiple-scattering suggest the formation of, primarily, large, multinuclear Co complexes or disordered hydroxide-like precipitates for Co sorbed to quartz. For Co sorption on rutile, the EXAFS spectra vary with changes in Co surface coverage and are distinctly different from Co/quartz spectra at all surface coverages examined. At low Co coverages, Co apparently sorbs directly to the rutile surface as mononuclear or small multinuclear complexes. This is indicated by two to three second-neighbor Ti atoms at two distinct distances, 2.95-2.99 A and 3.60-3.63 A, around sorbed Co. These Co-Ti distances are similar to Ti-Ti distances in the bulk rutile structure. The small number of Ti backscatterers indicates that Co has not diffused into the rutile structure but rather that Co occupies Ti-equivalent sites at the rutile surface. For high surface coverage samples equilibrated for eleven days, the average number of Co second-neighbor backscattering atoms increases, indicating larger-sized multinuclear complexes, but the EXAFS spectra do not resemble those of either the Co/quartz sorption samples or Co(OH)2(s). Analysis of these high coverage Co/rutile samples indicates Ti backscatterers at distances near those expected for anatase, a TiOz polymorph, suggesting a local geometry for Co and Ti octahedra at the rutile/water interface similar to that of the cation octahedra in anatase. We suggest that differences in Co sorption on quartz and rutile may be attributed to the availability of reactive surface sites on quartz and rutile which are structurally favorable for the sorption of octahedrally-coordinated Co 2+ . 1. INTRODUCTION

tems is essential for the understanding and the systematic quantification of these reactions in more complex natural systems consisting of heterogeneous soils and sediments in contact with a wide range of solution compositions. In both macroscopic studies of metal ion partitioning and spectroscopic studies of sorbed metal ions, a continuum from mononuclear adsorption, to multinuclear complex formation, and finally to surface precipitation with increasing metal ion surface coverage has been suggested (James et al., 1975; Farley et al., 1985; Sposito, 1986; Charlet and Manceau, 1992; O'Day et al., 1994a; Katz and Hayes, 1995a,b). This sequence of surface speciation is influenced by both the solution composition and the chemical and physical properties of the mineral substrate. At present, however, we lack quantitative understanding of the relative importance of solution and surface factors in influencing metal ion partitioning between aqueous and solid phases. Direct, in situ (i.e., with liquid water present) molecular-level spectroscopic investigations of sorbed species are required to determine their structure, stoichiometry, and uptake mechanism.

The partitioning of trace metal cations between aqueous solutions and solid oxide mineral phases may occur via different types of surface reactions. For example, adsorption of divalent transition metal ions involving partial loss of waters of hydration and complexation by surface oxygen or hydroxyl ligands ( "inner-sphere" complexation) has been proposed as a likely uptake mechanism at ambient temperatures and for relatively short equilibration times (hours to days). In contrast, some large, monovalent metal ions are thought to form "outer-sphere" surface complexes in which no waters of hydration are lost and the metal ion is held near the surface by relatively weak electrostatic forces. Although new information from a variety of molecular-level spectroscopic and microscopic techniques is beginning to emerge, the structure and composition of surface complexes, the nature of reactive sites on oxide surfaces, and the stoichiometry of surface reactions are not well known. Detailed knowledge of surface sorption reactions in well-constrained model sys2515

2516

P.A. O'Day et al.

X-ray absorption spectroscopy ( X A S ) was used to investigate the local molecular structure of in situ C o ( I I ) surface complexes sorbed to two pure mineral substrates, quartz ( a SiO2) and rutile (TiO2). Both of these minerals are simple oxides containing small, highly charged ( 4 + ) cations. Their mineral structures, cation coordination, and physical properties are, however, distinctly different. In quartz, Si is fourcoordinated by oxygen in a regular tetrahedron. In rutile, Ti is six-coordinated by oxygen in a distorted octahedron. Quartz is an insulating material with a very low, isotropic bulk static dielectric constant (e = 4.4-4.64; summarized in Shannon, 1993). Rutile is a semiconductor with one of the highest static dielectric constants measured for minerals and is extremely anisotropic (ellc = 170; e _1_ c = 8 6 - 8 9 ; Parker, 1961; Shannon, 1993). Although the sorption characteristics of metal ions onto these two minerals have been recognized as being dissimilar (e.g., James and Healy, 1972a,b), few studies have examined directly the local atomic structure of metal adions on these two surfaces using spectroscopic methods. In this study, absorption spectra were collected for samples equilibrated with solutions of different Co concentration and pH, resulting in different Co coverages on quartz and rutile surfaces. Synchrotron XAS is sensitive to the local ( < 6 A) atomic environment of a specific element, even at low concentrations (generally ppm) and in the presence of water (Brown et al., 1988). Recent spectroscopic (McBride et al., 1984; Bleam and McBride, 1985; Chisholm-Brause et al., 1990a; Manceau et al., 1992a; Waychunas et al., 1993; Junta and Hochella, 1994; Waite et al., 1994; O ' D a y et al., 1994a,b; Towle et al., 1995a,b) and theoretical (Sverjensky, 1993, 1994) studies have emphasized properties of the substrate in controlling surface complexation reactions. In this study, comparison of the E X A F S spectra of C o ( I I ) ions sorbed to quartz and rutile provides an illustration of how

mineral substrate may influence the local atomic coordination and binding of an adion on a mineral surface. 2. PREVIOUS STUDIES 2.1. Solution Studies

Uptake of Co(II) on quartz and rutile has been studied as a function of pH and initial metal ion concentration in batch partitioning experiments. Sorption studies of divalent transition metal ions (Me(II)) on quartz, different forms of amorphous SiO2, rutile, and anatase (a TiO2 polymorph) are summarized in Table 1. The general observations from these studies are: ( 1 ) Sorption of transition metal cations occurs at lower pH on rutile than on quartz or amorphous silica at similar total Me(II) concentrations, even though the pH point of zero net proton charge (pHpz,p~) of mille is much higher (5.5-5.8; Yates, 1975; Fokkink et al., 1989) than that of quartz (~2.5-2.9; Parks, 1965, 1975; Hunter, 1981 ). This observation was initially used as evidence by James and Healy (1972a,b) in support of a specific adsorption mechanism, rather than simple electrostatic attraction, for Co(H) uptake onto these surfaces (Fig. 1). (2) Uptake from solution occurs when the solution pH is within 1 - 2 pH units below the pH of precipitation of a hydroxide phase, or ~0.5-1 pH unit below significant hydrolysis of the metal ion in solution. (3) Sorption of Co is accompanied by surface charge reversals (as measured by electrokinetic methods) that, at higher surface sorption densities, resemble the surface charge behavior of Co(OH)2(s). This behavior has been attributed to the onset of precipitation or the formation of hydroxy-polymers which may be precursors to precipitates (Healy et al., 1968; James and Healy, 1972b; Matejevic and White, 1984). (4) Me(II) uptake is, in general, relatively insensitive to changes in the concentration of the background electrolyte, as shown in Fig. 1 for Co(II) uptake on quartz. This behavior has been used as indirect evidence for specific or innersphere adsorption of Me(II) on oxides (Hayes and Leckie, 1987). 2.2. Spectroscopic Studies

Spectroscopic investigations of metal ions sorbed at surfaces of amorphous silica suggest the formation of mononuclear and multinuclear complexes, or precipitation of surface phases, as solution pH and sorption density increases. A combined solution and X-ray pho-

TABLE 1. Summar~ of previous solution studies of divalent metal ion uptake on SiO~ and TiOg. Me(H)

Form

lMe(II)]total surface area/solution (m2/L) (Molality)

Me(II) somtion on SiOg: Co, Ni, Cu, Zn, Cd amorph. Pb quartz Co quartz Cu, Cd, Pb amorph, Zn quartz Cu, Zn, Cd, Pb quartz Ni quartz Cd amorph. Co, Ni, Cu, Zn, Cd amorph. Co quartz

10-2 10-4 10-6-10-4 n.r. 10-6-10-4 10-7-10-5 10-4.8 10-6-10-4 10"6-10-4 10-4-10-3

2 x 105 n.r. 75 n.r. 100 99 50 4480-5960 0.2-1.7 x 104 184-191

Co Cd Co Zn Cd Cd Co Cu

10-5-10-3 10-4 10-4 10-4 10-4 10-4-10-2 10-3-10-2 8.6 x 10-4

115 200 n.r. 110 150-200 3640-4840 1150 4470

n.r. = not reported

ruffle mtile n.r. anatase ruffle anatase, rutile rutile anatase

Reference

Dugger et al. (1964) Fuerstenau et al. (1970) James (1971); James and Healy (1972a) Schindler et al. (1976) James and MacNaughton (1977) Benjamin (1978); Benjamin and Leckie (1980) Theis and Richter (1980) Malati et al. (1982) Bye etal. (1982, 1983) O'Day (1992) James (1971); James and Healy (1972a) James et al. (1975) Tewari and Lee (1975) James and MacNaughton (1977) James et al. (1981) Malati et al. (1982) Chisholm-Brause (1991) Ludwig and Schindler (1995)

Sorption of

C o 2+ o n

100

O

,m

60

8 A

40

<>

)take on r

• I1.

20

0

'

2.0

~

~

"

4.0

6.0

"

|

8.0

"

"

"

100.0

pH

FIG. I. Co uptake from solution on powdered quartz and rutile as a function of pH compared with reported pH ranges of the point of zero net proton charge (pHp~op,.) for quartz (Parks, 1965, 1975; Hunter, 1981 ) and rutile ( Yates, 1975; Fokkink et al., 1989) surfaces (shaded bars). Co uptake on rutile (data from James, 1971): • [Co(II)]~o, = I.l × 10 5 M; solid/liq. = 115 m2/L; background electrolyte (B.E.) = 10 3 M KNO3; Co uptake on quartz (data from James, 1971): ~ [Co(ll)]t,, = 1.2 × 10 s M; solid/liq. = 75 m2/ L; B.E. = 10 3 M KNO3; Co uptake on quartz (data from O'Day, 1992): [Co(II)],,, = 3 × 10 4 M; solid/liq. = 185 ma/L; B.E. = O 0.01 M, [] 0.1 M, A 1.0 M NaNO~.

toelectron spectroscopy (XPS) study of Fe(II1) on porous silica postulated adsorption of monomeric species at low surface sorption densities (<0.18 ~mol m -~) and formation of low-molecular-weight Fe-oxy-hydroxide polymers at higher sorption densities (Anderson et al.. 1984). A similar conclusion was reached in studies of electron spin resonance (ESR) spectra of Mn ( II ) sorbed on silica gel (Hronsky et al., 1978, 1979). These studies suggested monodentate attachment of Mn to surface oxygens at low Mn 2+(aq) concentrations and lormation of hydrous Mn(II) clusters in the pores of silica gel at higher concentrations. In contrast to studies that suggested the formation of multinuclear species, other workers have interpreted ESR spectra of Mn(II) and Cu(ll) on silica gel as evidence for monomeric, hexaquocomplexes loosely bound to the surface (Burlamacchi and Martini, 1980; Martini, 1981 ). yon Zelewsky and Bemtgen (1982) attributed changes in the ESR spectra among aqueous Cu > , Cu 2+ sorbed on porous silica, and sorbed Cu > complexed by organic ligands as evidence for the formation of bidentate surface complexes. In their study of Cu ( II ) sorption on porous silica and other porous substrates, Bassetti et al. (1979) suggested that mobility of the aqueous ion as seen by ESR can be reduced by either chemisorption or by migration of the ion into narrow pores where adsorbed water has a different structure than bulk water. Unfortunately, ESR gives little information about the ligand structure around a sorbed ion to directly test this hypothesis. More direct structural information was provided by EXAFS and UV-vis reflectance studies of Ni ( II ) in the presence of a complexing agent (ammonia or ethylenediamine) sorbed onto or impregnated into a silica catalyst (Bonneviot et al., 1989a,b, 1990). These data were interpreted as evidence for a Ni-hydroxysilicate phase in the case of amine complexation and for isolated bidentate surface complexes of octahedrally-coordinated Ni in the case of the ethylenediamine complex. Recently, Charlet and Manceau (1994) have reinterpreted these data and suggested that the EXAFS results indicate "neoformation'" of non-stoichiometric hydrous silicates. It should be noted that all of the spectroscopic studies mentioned above used either silica gels or synthetic, precipitated or pyrogenic silicas as the

quartz and rutile

2517

substrate rather than ground, natural quartz as was used in this study. Furthermore, variations in sample preparation procedures for quartz can give rise to differences in surface behavior (Cases et al., 1985). It is not well known how surface atomic structures and metal ion bonding might differ among different forms or samples of SIO2. Few spectroscopic studies have been conducted on metal cations sorbed to rutile surfaces in situ or at ambient temperatures, but some inferences may be gained by considering studies on different forms of TiO2. Ottaviani et al. ( 1985 ) measured the uptake and electrophoretic mobility of aqueous Mn(II) and Cu(II) sorbed on amorphous TiO> followed by ESR measurements on dried samples. They concluded that Mn(II) bonds in the outer Helmholz plane without loss of waters of hydration, whereas Cu (II) bonds directly to the surface. They also suggested, based on ESR measurements, that water in the pores of the dried samples was structurally different from liquid water and was present in a "glassy" state. Luminescence decay spectra of Ru(II) bipyridene complexes sorbed to TiO2 (rutile and anatase) and SiO2 (amorphous) were studied by Kajiwara et al. (1982) and Hashimoto et al. (1988). These authors concluded that there were several adsorption sites for Ru complexes on the oxide surfaces and that electron-transfer rates from the complexes to the substrates differed among different adsorption sites. In a study by Ragai and Selim (1987), the amount of C o ( l l ) , Cu(ll), and Ni(II) uptake on different Titania gels, whose structures were controlled by preparing them at different pH and porosities, suggested that sorption occurs mainly at bridging hydroxyl sites. Using high-resolution transmission electron microscopy, Fendorf et al. (1992) noted surface precipitation of a possible Al-hydroxide on MnO2 (birnessite) but not on rutile at the same pH and AI solution concentration. In general, these studies suggest the specific adsorption of metal ions onto heterogeneous TiO2 surfaces and the lack of multinuclear complex formation or surface precipitation of metal hydroxides. Several other studies of divalent metal ion sorption on different forms of TiO2 provide some structural information, but they may not be directly comparable with this study because spectroscopic characterization was done after high-temperature processing. Hadjiivanov et al. (1991) used XPS, ESR, IR, and optical absorption measurements to study Co(II), Ni(II), and Cu(II) sorbed to anatase from ammonia solutions and calcined at 450°C. They concluded that the metal ions remained near the surface in a divalent, hexacoordihated state and that they interacted with Lewis acid sites (possibly surface Ti atoms) and not with surface hydroxyl groups. In a Fouriertransform infrared (FTIR) study of the anatase surface after reaction with water and heating in vacuum at 250°C, Van Veen (1989) found indirect evidence that Cd > sorption affected one of two principal OH stretching frequencies, the "acid" rather than the "basic" hydroxyl site. Like the low-temperature studies, the high-temperature work also suggests interaction of metal ions with specific sites on TiO2 surfaces. Preliminary EXAFS analysis of one Co/mtile sample discussed in this study (surface coverage ( F ) = 1.38 ~zmol m -2 ) was reported in Chisholm-Brause et al. ( 1990a); initial results for two Co/quartz (samples " A " and " B " ) were reported in O'Day et al. ( 1991 ). In these studies, we suggested the formation of inner-sphere, multinuclear Co complexes on different oxide surfaces. We have re-analyzed these three sorption samples and ten new samples in this study using much better experimental and theoretical standards than were available at the time of the initial publications. In particular, we have focused on the analysis of backscattering from second-neighbor atoms around sorbed Co. This provides structural information about surface sorption sites for inner-sphere complexes and, for multinuclear complexes and surface precipitates, it provides information about their size and degree of order. Although the interpretations of the original studies are still largely valid, this study reports a much more detailed picture of the local atomic structure around the sorbed metal ion and how this structure changes with changes in surface coverage and equilibration time. 3. E X P E R I M E N T A L M E T H O D S 3.1. Solids The quartz used in this study was Min-U-Sil 30 obtained from Pennsylvania Glass Sand Corporation. It was prepared by the manu-

2518

P.A. O'Day etal.

facturer by grinding natural quartz sand (from Berkeley Springs, WV) in a silica-lined ball mill using silica pebbles and then sizefractioned in an air-cyclone separator. A typical chemical analysis provided by the manufacturer is: sin2, 99.7%; Fe203, 0.023%; A1203, 0.101%; Tin2, 0.019%; Can and MgO, trace; average particle size is 8.8/zm. Due to the presence of trace impurities and fine particles that can affect sorption properties, significant treatment of commercial quartz of this type is necessary (D. Kent, pers. commun.). The ground material obtained from the manufacturer was suspended in deionized water and allowed to settle through 12 cm of water for 24 h to remove particles of <2 /zm (that might be amorphous). The supernatent solution containing fine particles was removed and the material resuspended. This was repeated twentyfour times until the supernatent solution was visibly clear. The >2 #m material was then washed twice in 1.6 N HNO3 for 12 h at 65°C and then five times in deionized water, ten times in 0.1 M NaC1, and again with deionized water (12-24 h settling time between washes). Evidence of algal activity indicated that more severe treatment was necessary. The material was heated in air at 450°C for 48 h and treated with 4 N HC1 at 65°C for 4 h. It was then washed in 0.1 M NaCI, centrifuged, and the supernatent solution was removed. Washing and centrifuging were repeated twenty-one times with NaC1 and five times with deionized water. Finally, the quartz was dried for several days at 65°C and stored in glass containers. The surface area of the cleaned, dried material measured by N2 BET method (Micromeritics BET Accusorb 2100E) was 0.77 m2 g-~. Two different lots of synthetic rutile were used in this study; both were synthesized by the same method (hydrolysis of TIC14). Unlike ground quartz, synthetic rutile is very pure and little or no cleaning is required. For Co sorption samples at the two lowest surface coverages (F = 0.63 and 1.38/zmol m z), the rutile was obtained from British Titan Products (courtesy of R. O. James, Eastman Kodak). This material was washed for 24 h with deionized water in Soxhlet extraction apparatus to remove trace chloride (James, 1971). The surface area measured by N2 BET method was 23 m 2 g-~. Only a limited supply of this high surface area rutile was available, so a different ruffle, supplied by Dupont (R-900-CD; courtesy of R. E. Johnson, Dupont), was used for the three samples at higher surface coverage. No pretreatment was done on this material and analysis by XPS indicated no significant surface impurities. The Nz BET surface area was measured at 6.07 nag g-~. For both rutiles, their identity and crystallinity were confirmed by X-ray diffraction and electron microscopy. Both lots of rutile appear morphologically similar and contain only minor anatase ( < 1% of the bulk sample).

3.2. Sample Preparation for EXAFS Sorption samples were prepared by equilibration of individual batches of quartz and ruffle with NaNO3/Co(NO3)2 solutions in polypropylene tubes at surface area to solution ratios and aqueous concentrations shown in Table 2. Sorption experiments were carried out in a N2 atmosphere with N2-sparged aqueous solutions at ambient conditions (T = 22 ° _+ 3°C). Solutions were prepared with ACS reagent-grade salts and double-deionized water. Solids were mixed for 6-15 h with 0.1 M NaNOs (pH ~ 6 for quartz; pH ~ 3.5 for rutile) to wet and hydroxylate the solid. Aqueous Co(NO3)2 spiked with carrier-free, radioactive 57Co (half-life = 270 days) was added to the suspensions at Co(II) concentrations ([Co(II)]tot) shown in Table 2 and equilibrated while rotating for 6-24 h. Solution pH was then raised by adding small amounts of NaOH to achieve uptake of Co from solution. Base was added dropwise while stirring the suspension to minimize local mixing zones of high pH where oversaturation with respect to solid hydroxide phases might occur. The suspensions were then equilibrated while rotating for the times shown in Table 2, after which pH was measured. A small aliquot (0.3-0.5 mL) of supernatent solution was taken by either filtration using a 0.45 #m nylon membrane filter or after centrifuging the suspension. Cobalt activity from gamma radiation was measured in the supernatent solution with a Tracor Northern (TN-1750) Multichannel Analyzer. Blank Co solutions were prepared in exactly the same manner as the sorption samples but with no solid present and no adjustment of pH with acid or base. Uptake was calculated from the difference in counts between the sample and the blank solution according to

I %uptake= 100x

sample cts (s - ~ g ) ) 1-

1

-bkgrd. cts(s - ) g - ~ ) j blank cts (s -~ g-~) - bkgrd, cts ( s - ~ g - 1)

(11

where bkgrd, cts are the measured background counts for an empty vial. Surface sorption density (F) is calculated according to F(mol m 2) = [Co(II)]to,(M) x (% uptake) solid/liquid (m 2 L -~) '

(2)

where [Co(II)]tot is the total concentration of aqueous Co and the solid/liquid ratio is calculated from the BET surface area per liter of solution. Error in pH measurements was estimated by replicate samples and measurements to be _+0.02 pH units. Average relative statistical error from counting of radioactive Co disintegrations for a single measurement was estimated at _+2%. For XAS data collection, the suspensions were centrifuged and, under N2, excess supernatent solution was removed. The samples were loaded as wet pastes into Teflon sample holders and sealed with Mylar windows for XAS measurement. For most XAS samples, Co uptake from solution of >90% was achieved while maintaining undersaturation of the solution with respect to precipitation of Co-hydroxide phases. This minimizes the X-ray absorption signal from Co 2+(aq) present in the small amount of residual supernatent solution retained in the XAS samples to keep them wet during data collection. Several samples, however, had Co uptake from 43-84% (Table 2). In these samples, we estimate that the contribution to the EXAFS signal from Co in the residual aqueous solution is less than 10% of the total Co fluorescence signal. Three quartz samples were equilibrated with solutions in which the thermodynamic solubility of all three known forms of Co(OH)z(S) ("blue," "pink, active," and "pink, inactive"; Feitknecht and Schindler, 1963; Baes and Mesmer, 1976) was purposely exceeded as pH was raised during the addition of NaOH. Two of the quartz sorption samples, 1-" = 9.21 and 7.41 #tool m -2, were equilibrated for 24 h at an oversaturated pH, then pH was lowered (by addition of HNO3) to undersaturated conditions and the samples were reequilibrated for the times shown in Table 2.

3.3. XAS Data Collection and Analysis X-ray absorption spectra were collected on wiggler beam lines IV-1 and IV-3 at the Stanford Synchrotron Radiation Laboratory (SSRL), Stanford, CA, and on beam lines XII-A and X8-C at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory, NY. Beam current at SSRL varied from 20 to 90 mA at 3 GeV and the magnetic field of the wiggler was 18 kG. At NSLS, beam current was 100-210 mA at 2.5 GeV. Either S i ( l l l ) or

T A B L E 2. Experimental conditions for E X A F S sorption samples. F(I) (10 -6 mol m "2)

solid/liq. (m2L -1)

[Co(//)]tot [Co(II)]final (10 -3 M) (10- 4 M)

pHfinal

Equil. time

% Co uptake

10.70 9.84 8.99 7.04 6.78 7.76 8.50 8.74

24 24 21 42 54 47 36 24

h h d. h h h h h

>98.0 >98.0 >98.0 84.2 68.5 96.2 99.1 98.6

7.64 7.49 7.89 6.72 5.29

11 11 23 24 23

d. d. h h h

96.2 94.1 97,0 95.1 43. I

C o 0 I ) sorption on quartz: A(21 B(21 C(2) 9.11(3) 7.41(3) 5.21 1.61 0.77

192 192 185 185 185 185 185 192

1.90 1.20 2.00 2.00 2.00 1.00 0.30 0,15

<0.38 <0.24 <0.40 3.17 6.30 0.38 0.03 0.02

3.00 1.50 0.50 1.67 1.67

1.15 0.88 0.15 0.82 9.50

Co(lI) sorption on rutile: 9.51 4.65 1.60 1.38 0.63

304 304 304 1150 1150

( I )Surface sorption density given by Eqn, (2) Background electrolyte for all samples was 0.10 M NaNO3 (2)Sorption sample equilibrated with solutions oversaturated with respect to solubilites reported tot all stable and metastable forms of Co(OH)2(s) (Feitknecht and Sehindler, 1963; Baes and Mesmer, 19761. All other samples equilibrated with solutions under'saturated with respect to Co(OH)2(s) phases. (3) Initial equilibration for 24 h with solutions oversaturated with respect to Co(OH)2(s) phases; pH adjusted with addition of acid to undersaturated conditions and equilibrated for the time shown.

Sorption of Co g+ on quartz and rutile Si (220) monochromator crystals were used with an unfocused beam. Rejection of higher-order harmonic reflections was achieved by detuning the monochromator such that the maximum incoming beam flux was reduced by 30-50%. Beam energy was calibrated by assigning the first inflection on the absorption edge of metallic Co foil to an energy of 7709.3 eV. There was no evidence for significant oxidation of Co(II) to Co(III) in any of the sorption samples or precipitates. Oxidation would be indicated by a 2 - 5 eV increase in the position of the Co absorption edge (Manceau et al., 1992b) and/ or discoloration of the sample. Fluorescence spectra for wet sorption samples were collected at 45 ° to the incident beam using, for most samples, a Stern-Healdtype detector (Lytle et al., 1984) with Soller slits and an Fe filter to reduce background scattering and fluorescence. Spectra for two of the ruffle sorption samples (F = 4.65 and 9.51 #tool m 2) were collected with a five-element Si array detector; due to electronic problems, however, only two of the five channels were usable. Spectra for the rutile sorption sample at F = 1.60 #tool m 2 were collected using a thirteen-element Ge array detector and AI filter to reduce background fluorescence. Transmission spectra were collected for the Co(OH)2(s) model compounds using gas-filled ion chambers. Multiple scans (three to sixteen, depending on Co concentration) were collected and averaged for each sample to improve the signal-to-noise ratio. In general, the background noise level is noticeably higher in the Co/rutile spectra than in the Co/quartz spectra. This noise results from secondary absorption and scattering processes, and these effects are more pronounced for elements such as Ti and Fe than for elements like Si. Although filters were used to maximize signal-to-noise in both cases, much of the noise in the futile spectra is probably a result of secondary absorption of Co fluorescence by Ti in rutile ( "fluorescence quenching" ) and perhaps increased inelastic scattering within the sample. Data collection and analysis procedures are described in detail in O'Day et al. (1994a,c). Briefly, numerical results were extracted from the EXAFS spectra using a curved-wave formalism and a single-scattering approximation (reviewed in Stern, 1988) implemented in the computer code EXAFSPAK (G. George, SSRL). Background below the absorption edge was estimated by a linear fit throught the pre-edge region for all samples except for the futile sample collected with the Ge array detector, for which a Gaussian fit was used. Background above the edge was determined by fitting a cubic spline of three or four points. Spectra were normalized using the edge-step height near the absorption edge and extrapolated through the EXAFS using a Victoreen polynomial and tabulated McMaster coefficients (McMaster et al., 1969; Teo, 1986; Brown et al., 1988). A value of 7725 eV was used for Eo (the energy at which k, the photoelectron wave vector, is zero). Spectra were weighted by k 3 to compensate for damping of oscillations at high k. Normalized, background-subtracted EXAFS for reference and unknown spectra were filtered over similar k-ranges (k ~ 3 to 12-13 A, ~) and Fourier-transformed (square windowing) to produce radial structure functions (RSFs) that isolate frequency correlations between the central absorber atom and backscanering atoms as a function of distance (R). The RSFs are shown here uncorrected for phase shifts of the backscattering atoms. This results in RSF peak positions of backscattering atoms at apparent distances which are shorter than the true distances by about 0.4-0.5 A. Interatomic distances are corrected for phase shift during the fitting procedure described next. Nonlinear least-squares methods were used to fit unknown spectra to reference EXAFS phase-shift and amplitude functions. Reference functions were taken from experimental data or theoretically calculated using the theoretical code FEFE6 (Mustre de Leon et al., 1991; Rehr et al., 1992; Rehr, 1993). Fits using either theoretical or experimental reference functions agreed within experimental error in all cases for which good experimental reference functions were obtainable (O'Day et al., 1994c). Because of the lack of good experimental reference compounds, especially lot second-neighbor backscattering atoms, we relied on FEFF6 theoretical functions for Co-Si and CoTi pair correlations and for the analysis of multiple scattering. To generate these reference functions, mineral structures were used to calculate atomic clusters of reasonable geometry around Co atoms by, for example, substituting Co for Ti in the rutile structure. Individual scattering paths at different distances (e.g., Co-O, Co-Ti) were extracted from the cluster calculation and used as reference functions

2519

for unknown spectra. Typically for unknown spectra, R, N, and ae, a Debye-Waller term that accounts for thermal and static disorder, were treated as adjustable parameters for each set of backscattering atoms at the same distance. The difference in Eo(AEo) between the reference functions (either experimental or theoretical) and the unknown spectrum was treated as a single adjustable parameter for all sets of backscattering atoms for each sample (O'Day et al., 1994c). Reference functions were fit initially to filtered EXAFS spectra of individual peaks in the RSFs to determine backscatterer identities and interatomic distances. Final fits were done on the full normalized spectra to account for any overlap in backscattering functions. Because of the large number of adjustable parameters when fitting normalized spectra with more than two sets of backscattering atoms, all possible adjustable parameters could not be refined simultaneously. For the final fits, sets of parameters were alternately varied and fixed until the least-squares function was minimized, or in some cases, one of the adjustable parameters (usually N or cT~') was fixed at a value estimated from known compounds and all other parameters varied. Estimates of maximum errors in the EXAFS analysis were obtained in a previous study which compared experimental EXAFS spectra of known Fe, Co, and Ni crystalline compounds and theoretical EXAFS functions generated by FEFF (O' Day et al., 1994c ); These errors are: Co-O, R _+ 0.02 A, N _+ 20%; Co-Co, R + 0.02 A, N _+ 15%; CoSi or Co-Ti, R _+ 0.02 A, N _+ 25%; error in ~2 for all absorberbackscanerer pairs is estimated at _+20 30%. 4. RESULTS

4.1. EXAFS Results: Co(II)/Quartz Cobalt sorption on quartz was studied by X A S at surface coverages ( F ) ranging from 0.77 to 9.21 ~mol m 2. Five sorption samples were equilibrated with solutions undersaturated with respect to C o ( O H ) 2 ( s ) and three samples were equilibrated with oversaturated solutions, one of which was aged for twenty-one days (Table 2). The normalized, background-subtracted E X A F S spectra and RSFs are compared with those of a dry, crystalline C o ( O H ) 2 ( s ) model compound in Fig. 2. In the normalized E X A F S spectra, the position and shape of the E X A F S oscillations are very similar for all quartz samples regardless of surface coverages and equilibration time, although the noise level is higher in the lowest F samples. Comparison of the E X A F S results among the sorption samples and the C o ( O H ) ~ ( s ) reference compound (Table 3) indicates the following: 1 ) The first peak in the RSFs corresponds to backscattering from O atoms ligating Co. In all spectra, the oxygen coordination around Co is the same and nearly identical to that of Co in Co (OH)2 ( s ) ( N = 6, R = 2.097 A ) , consisting of six (_+20%) O atoms at 2 . 0 6 - 2 . 1 1 (_+0.02) A and indicating octahedral coordination of Co. There is no evidence in either the E X A F S or the pre-edge structure for significant distortion of the oxygen octahedron around Co (ChisholmBrause et al., 1990a; O ' D a y et al., 1994b). 2) The second peak at R = 2 . 2 - 3 . 2 ,~ (uncorrected for phase shift) corresponds to backscattering from secondneighbor atoms beyond the O coordination shell. In C o ( O H ) 2 ( s ) , this peak corresponds to six Co atoms at R = 3.173 A. In the sorption samples, E X A F S oscillations from k ,~ 5 to 10 are shifted to higher k and have a slightly higher frequency compared with the Co (OH)2 ( s ) normalized spectrum. Also note that the sorption spectra lack " s h o u l d e r s " or " b e a t s " on the major peaks that are prominent in the E X A F S of C o ( O H ) 2 ( s ) (Fig. 2). The higher frequency in k-space (manifested as a shift of the peaks in the RSFs to lower R ) indicates a contraction of interatomic

2520

P.A. O'Day et al. 1n

t-e3

~4 v

3

4

5

6k (/~ -17)

8

9

10

11

v0

1

2 R(/~)3 4

5

6

FIG. 2. Normalized, background-subtracted EXAFS spectra (weighted by k 3) and Fourier transforms (uncorrected for backscatter phase shift) of Co sorbed to quartz at surface sorption densities shown (in #tool m -2) compared to the spectrum for dry, crystalline Co(OH)2(s) model compound (top spectrum). Sorption samples A, B, and C are samples equilibrated with aqueous Co solutions oversaturated with respect to the reported thermodynamic solubilities of all forms of Co(OH)z(S) (see Table 2). Samples with Co surface coverages of 9.2 and 7.5 #tool m -2 were initially equilibrated with oversaturated Co solutions; pH was adjusted (by addition of acid) to undersaturated conditions and re-equilibrated (Table 2).

distances in the sorption samples relative to Co(OH)2(s). The numerical results from least-squares fitting of the second-coordination shell are discussed below. 3) The peak at R = 5.5-6.3 ,~ results from multiple scattering among equidistant Co atoms whose total path length is exactly twice that of the second RSF peak. Note the expected contraction in distance (lower peak position in the RSFs) for the sorption samples compared to Co(OH)2(s). Analysis of this peak is also discussed below.

4.1.1. Analysis of the second coordination shell The filtered EXAFS of the second RSF peak of the sorption samples were initially fit assuming only backscattering from Co atoms as in Co(OH)z(s). Fitted Co-Co distances in the sorption samples are shorter (R = 3.11-3.12 ,~) than Co-Co distances in C o ( O H ) 2 ( S ) ( R = 3.173 A ) . This result is similar to that observed for Co sorption on kaolinite (O'Day et al., 1994a,b) and on alumina (Chisholm-Brause et al., 1990a; Towle et al., 1995b). Although this fit model accounted for most of the amplitude of the peak, these fits produced higher numbers of Co backscatterers (N > 7) than expected for octahedrally-coordinated Co is in a closestpacked structure (N = 6). In addition, there was a slight

mismatch in the position of the peak in R-space in the RSF when the fit was optimized in k-space and six Co atoms were assumed in the model (Fig. 3). Detailed EXAFS analysis of second-neighbor backscattering in known mineral reference compounds has shown that, in some cases, the presence of a second, weaker backscatterer is indicated by higher-thanexpected coordination numbers and a slight phase mismatch in the least-squares fit when only one backscatterer is assumed (for details, see O'Day et al., 1994c). In light of these results, we tested a series of structural models of possible sorption geometries by calculating theoretical EXAFS with FEFF6 and comparing them to the experimental data. Evidence from solution and spectroscopic studies discussed previously suggests that Co may bond to quartz as an inner-sphere complex. In this case, distances from Co to Si atoms in the substrate should be close enough ( < 4 A) to be detected by EXAFS. To test this bonding geometry, normalized spectra were fit assuming one shell of six Co atoms and one shell containing Si atoms. Theoretical EXAFS for Co-Si correlations were calculated with FEFF6 using hypothetical atomic clusters of Co substituted into the structures of two compounds, lizardite (Mg3Si205 (OH)a, Co substituted for Mg; Mellini and Zanazzi, 1987) and LiFeSi206 (Co substituted for Fe; Clark et al., 1969). When Co is

Sorption of Co 2+ on quartz and rutile

2521

TABLE 3. Results of EXAFS anal),ses of Co sorbed to quartz and rutile.

Co~ll) sorption on quartz: F (10-6mol m "2) A0) B0) CO) 9.21 7.41 5.21 1.61 0.77

Co-O N(2) cr2(,~2)

R(.~) 2.07 2.06 2.08 2.08 2.08 2.08 2.09 2.11

6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0

0.0115 0.0097 0.0065 0.0070 0.0074 0.0068 0.0092 0.0073

R(A)

Co-Si N ¢32(,~2)(2)

3.37 3.38 3.40 3.41 3.40 3.40 3.42 3.46

1.8 2.7 3.1 2.7 2.2 3.0 1.9 1.1

R(.i.)

Co-Co N(2) 0-2(A2)

0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005

3.12 3.11 3.12 3.12 3.12 3.12 3.12 3.11

6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0

G2(.A2)

R(,~)

N

G2(A 2)

3.10

3.9-4.0

0.0073-0.0076

-0.5

3.13

3.7-5.0

0.0073-0.0082

-3.1

3.12

1.3-1.5

0.0056-0.0062

-5.9

0.0074 0.0075 0.0072 0.0065 0.0069 0.0068 0.0074 0.0064

AEo(eV)(3) 1.5 -0.2 0.8 1.5 1.7 1.6 1.7 1.9

Co¢ll) sorDtion on rutile: F

(10-6moi m "2)

17o-O

Co-Ti

R(,~)

N(2)

02(,~ 2)

9.51

2.06

6.0

0.0074

4.65

2.07

6.0

0.0094

1.60 1.38

2.08

6.0

0.0075

0.63

2.07

6.0

0.0047

R(A.)

N

Co-Co

3.01 2.0 0.010(2) 3.81 2.7-2.8 0.015(2) 3.01-3.05 2.4-2.9 0.010(2) 3.83 1.9-2.9 0.015(2) -- no quantitative analysis -2.99 2.2 0.012 3.63 3.5-5.5 0.014-0.021 2.95 1.4 0.010 3.60 3.6-4.2 0.0071-0.0085

AEo(eV)(3)

-6.9

(1)Thermodynamicsolubilityof Co(OH)2(s)exceeded. (2)Fixedparameter. (3)AEofor FEFF6theoreticalreferencefunctions.AEo usingexperimentalreferencefunctionsfor Co-Oand Co-Covariedbetween-2.0and -4.0eV for quartz and -3 to -5 eV for rutile.

substituted and treated as the central absorber in both of these compounds, second-neighbor Si atoms (Co-Si = 3.287 for Co-substituted lizardite; Co-Si = 3.208-3.345 ,~ for Co-substituted LiFeSi206) are in close proximity to secondneighbor Co atoms (Co-Co = 3.065-3.080 ,~ for lizardite and 3.177 ,~ for LiFeSi206) around a central Co atom. These distances, and the local electronic potentials of the bonding environment, are similar to those expected for Co sorbed as an inner-sphere, multinuclear complex on quartz and they provide a reasonable starting model. Nonlinear least-squares fits were done using both normalized and filtered (back-transformed) data in which different sets of variables were alternately fixed and varied. Because of the high degree of correlation between N and o-2 and the low amplitude from Si backscattering, it was found that these two parameters could not be refined simultaneously for the Co-Si path. In addition, the sine-wave oscillations of Co-Si scattering at 3.2-3.4 ,~ are nearly in phase with that of CoCo at ~3.1 ,~ in the k-range of 6 - 9 ,~-1 (Fig. 3). The weak Si backscattering amplitude and the overlap with the CoCo backscattering function make it impossible to resolve uniquely the presence of Si in the second shell unless a value for the Debye-Waller factor (o-2) for Co-Si is assumed. A value of o-2 = 0.005 ~2 for Co-Si scattering was tested in the fit model. This value is based on a previous EXAFS analysis of reference compounds containing Fe, Co, and Ni as the central absorber, O or F as first-shell atomic backscatters, and Si and metal ions as second-neighbor backscatterers (O'Day et al., 1994c). As shown in Fig. 3, a model based

on these assumptions gives a good fit to the experimental data, although some EXAFS amplitude in the k-range of 4 6 ,~-1 is not accounted for. Owing to the inverse correlation between N and o-e, increasing o-2 from 0.005 ,~2 to 0.01 ,~2 for Co-Si results in a corresponding decrease in N for Si, but does not significantly change interatomic distances. The data reported in Table 3 are the best fits to the normalized EXAFS from a model in which N was fixed at six for the Co-O and Co-Co paths, o-2 was fixed at 0.005 ,~a for the Co-Si path, and all other parameters were varied. In all fits based on this model, Co-Si distances ranged from 3.37-3.46 and the number of Si backscatterers was approximately 1-3 (Table 3). Although this result is consistent with a small fraction of Co atoms bonded directly to quartz tetrahedra (discussed below), it should be emphasized that this model assumes the presence of Si backscattering, which is weak compared with Co backscattering and is masked by the overlapping sine-wave oscillations of Co-Co and Co-Si paths. It must also be emphasized that the experimental EXAFS is a weighted average of all Co bonding environments in the sample. The fitted results reflect this averaging, and it may account for the range of Co-Si distances (3.37-3.46 ,~) derived from the least-squares analysis. Given these observations, it is difficult to assess realistic errors for Co-Si scattering paths, but they are likely to be quite high, especially for the number of Si backscatterers (probably >25%). The statistical significance of the improvement in the EXAFS fit when another set of variables is added to the model cannot be numerically

2522

P.A. O'Day et al. 10 -

I

I

I

l I

.

I

I

I

I

t-

'

~ -10

-15

I

ii J

I

I

°-c°

Co-O

!

-5 t',',

I

I t.Co-Si

,31

Z

MS

L. Co-O

-20 I

-25

-30

MS

I

4

I

5

I

6

I

I

7 8 k ( k -1 )

I

9

I

10

I

11

12

1

2

3

4

5

6

R(A)

FIG. 3. Normalized EXAFS spectra (solid line) and nonlinear least squares fit (dashed line) of Co/quartz sorption sample at surface coverage of 5.21 #mol m 2. Left: Best fit requiring 6 first-neighbor O atoms, six second-neighbor Co atoms, and Si atoms (or2 fixed at 0.005 A J) in the fit model (see Table 3). Deconvolution of the fit into x(k) components for each set of atoms is shown at the bottom. Right: Fourier transform spectrum shows the deconvolution of the fit (dashed line) into contributions from O and Co backscanerers (dotted line) with no Si assumed in the fit (bottom) and with Si required in the fit (top). MS: Multiple-scattering peak. determined because we presently lack robust statistical tests for correlated variables in nonlinear least-squares fitting. Therefore, it is necessary to rely on empirical comparisons with known reference compounds. Studies to date (e.g., O'Day et al., 1994c; Thompson et al., 1995) suggest that the ability to distinguish multiple sets of backscattering atoms beyond the first coordination shell varies widely among different systems and that comparisons with unknown systems must be viewed with caution. In general, our EXAFS results indicate that the local atomic environment around Co sorbed to quartz is similar, but not identical, to that of Co in Co(OH)2(s). This suggests the formation of, primarily, large, multinuclear complexes or a hydroxide-like precipitate. Shortening of Co-Co secondneighbor distances compared to distances in Co(OH)2(s) is similar that observed in other Co sorption systems. Some of the spectral differences between the Co/quartz sorption samples and Co(OH)2(s) may be explained by the presence of a small average number of Si atoms contributing to the EXAFS spectrum. Due to the weak amplitude of this scattering, however, we cannot definitively demonstrate the presence of Si in the second coordination shell around sorbed Co and other scattering contributions may account for the differences between the sorption sample spectra and the Co(OH)2(s) spectrum.

4.1.2. Analysis of rnultiple scattering The peak in the RSFs at R = 5.5-6.3 ,~ (Figs. 2, 3) was analyzed using FEFF assuming multiple-scattering among

Co atoms at ~3 and 6 A. This analysis is a semiquantitative comparison of EXAFS amplitudes resulting from linear multiple scattering among Co atoms at ~ 3 and 6 ,~ in the sorption samples relative to the same multiple-scattering paths in crystalline Co(OH)2(s) (see O'Day et al., 1994a, for details of the analysis). Multiple scattering of this type is strong enough to contribute to the EXAFS signal and can provide information about backscattering atoms beyond the nearest coordination shells, in this case, out to ~ 6 A (O'Day et al., 1994c). Briefly, this analysis (1) assumes a local atomic structure in the sorption samples that is similar to the planar structure of Co(OH)2(s), (2) assumes similar multiple-scattering paths in the sorption samples and C o ( O H ) f f s ) , and (3) compares the amplitudes of the EXAFS multiple-scattering functions, both theoretical (generated by FEFF) and experimental, in crystalline Co (OH)2 (s) and the sorption samples and expresses the differences relative to 100% amplitude assumed for Co(OH)2(s). Analysis of the Co multiple-scattering EXAFS (using filtered data) indicates that scattering amplitudes are nearly the same for second-neighbor (3.10-3.12 ,~) Co atoms in the quartz sorption samples and Co(OH)2(s). Scattering from distant Co atoms (6.22-6.24 ,~), however, is reduced significantly in the sorption samples relative to that in Co(OH)2(s). Specifically, filtered scattering amplitudes from multiple scattering were reduced by only about 0 - 2 0 % for Co atoms at 3.10-3.12 A, but by 4 0 - 7 0 % for Co atoms at 6.22-6.24 in the sorption samples relative to amplitudes observed for Co(OH)2(s). These amplitude decreases may result from

Sorption of Co > on quartz and ruffle

2523

70

60

50

40

o%

30

20

10

-t03

4

5

6

7_1 8 k(/~ )

9

10

ll

12 R(A)

FIG. 4. Normalized, background-subtractedEXAFS spectra (weighted by k 3) and Fourier transforms (uncorrected for backscatterer phase shift) of Co sorbed to ruffle at surface sorption densities shown (in #mol m-2) compared to the spectrum for dry, crystalline Co(OH)2(s) model compound (top spectrum) and to the spectrum of a 12 mM aqueous solution of Co 2+ (bottom spectrum). Samples with Co surface coverages of 0.63 and 1.38 #mol m-2 and those with coverages of 1.60, 4.65, and 9.51 #tool m-2 were equilibrated with two different lots of ruffle (see text).

lower numbers of Co backscatterers, increased static disorder in the hydroxide-like structure, and/or nonlinear scattering paths among Co atoms. These results are similar to those found for Co sorbed to kaolinite at high surface coverages (O'Day et al., 1994a). No systematic variations in the EXAFS spectra were noted as a function of changes in surface loading of Co on quartz. Most of the variation in the leastsquares analyses was a consequence of noisier spectra for sorption samples at low coverage.

4.2. EXAFS Results: Co(II)/Rutile Normalized EXAFS and RSFs of Co sorbed to rutile are compared in Fig. 4 to an aqueous solution of Co 2+ (12 mM, no solid present) and to the Co(OH)2(s) reference compound. Although the noise level is higher in the Co/ rutile samples, the pattern of oscillations in the normalized spectra is distinctly different from either the quartz spectra or the Co(OH)2(s) spectrum (compare Figs. 2, 4). This indicates differences in the local atomic structure around Co sorbed on rutile. Least-squares analyses of the normalized spectra indicate that the first-shell Co coordination is similar to Co coordination on quartz and in Co(OH)2(s), consisting of about six O atoms at 2.07-2.08 A. The RSFs for the Co/

futile samples at low coverages, however, show two secondneighbor peaks, rather than one peak, at R ~ 2-3.5 ,~ (peaks in the RSFs at higher R are spurious and result from spectral noise). In contrast to quartz, there are distinct changes in the EXAFS and RSFs with increasing Co surface coverages on rutile. At the highest surface coverages, the EXAFS oscillations do not resemble those of either Co sorbed to quartz or Co(OH)2(S). The differences in the EXAFS spectra among the Co/ rutile sorption samples arise from differences in R, N, and O-2 among at least two and probably three different sets of second-neighbor backscattering atoms. Due to the complexity of the spectral analyses and the high noise level in many of the samples, all possible adjustable parameters could not be refined simultaneously in a least-squares fit. The approach used here was to propose different, structurally reasonable atomic clusters around sorbed Co as the initial fit model and then to adjust sets of parameters individually until a good match with both the normalized EXAFS spectrum and the RSF was found. This approach was fairly robust for distance determination; i.e., different starting models converged to the same answer. Because of the high degree of correlation between N and cr2 and the overlap in the sine-wave backscattering functions of different atoms, however, it was not possi-

2524

P.A. O'Day et al.

ble to refine N and o-2 simultaneously in least-squares fitting. Therefore, we report ranges of N and o 2 in Table 3 that represent equally good fits to the data.

4.2.1. Analysis of the second coordination shell At the lowest surface coverage (F = 0.63 #tool m 2), EXAFS spectra indicate average second-neighbor coordination around Co of about 1.4 Ti atoms at 2.95 A and 3.64.2 Ti atoms at 3.60 ,~, indicating inner-sphere complexation of Co at the rutile surface (Table 3). Assuming the presence of second-neighbor Co (Nco -> 1 ) in the fit model results in a significantly worse fit. Backscattering from fewer than one second-neighbor Co, however, cannot be distinguished in the EXAFS spectrum of this sample. At the next highest surface coverage (F = 1.38 #tool m 2), the presence of a small number (Nco = 1.3-1.5 ) of second-neighbor Co atoms at 3.12 ,~ in addition to the Ti backscatterers is required to accurately reproduce the normalized EXAFS and RSF (Fig. 5a) and the presence of second-neighbor Co is clearly indicated at this surface coverage. Compared to the quartz samples at similar coverages (Table 3), the average number of Co second-neighbor backscatterers at the two lowest coverages on rutile is much smaller (Nco = 0 - 1 . 5 ) . Also, two distinct Co-Ti distances, rather than one, are clearly present in the Co/rutile spectra. The Co/rutile sorption samples at three higher surface coverages (1.60, 4.65, and 9.51 #mol m 2) were prepared with a different synthetic rutile than the two lower coverage samples. Due to experimental complications, quantitative data analysis of the sample at F = 1.60 #mol m -2 over the full k-range (3 to 12-13 A J) could not be carried out. Qualitatively, the EXAFS oscillations of this sample are intermediate between high and low coverage samples, but are more similar to the higher coverage samples (4.65 and 9.51 # m o l m 2; Fig. 4). Two significant differences in the EXAFS results are apparent in the two highest coverage samples compared with the two lowest samples (Table 3): (1) the average number of second-neighbor Co atoms is higher, although it does not reach Nco = 6, the number expected for the Co(OH)z(S) structure; (2) Co-Ti interatomic distances are longer than in the two lowest coverage samples (Fig. 5b). The accurate determination of N and a 2 in these samples was complicated by amplitude cancellation in the backscattering sine-wave functions among the different sets of Ti and Co backscatterers. In addition, we cannot preclude the possibility of multiple Co sorption sites and thus, multiple sets of Ti backscatterers. The Co-Co distance, however, is similar for all rutile samples and suggests edge-shared Co octahedra as seen for Co sorbed to quartz. In all of the Co/ rutile samples, there is no evidence for significant linear multiple-scattering in the EXAFS spectra as was observed in the Co/quartz samples and Co(OH)2(s). These results indicate the formation of mononuclear and multinuclear inner-sphere Co complexes on rutile, but not large hydroxidelike multinuclear complexes or precipitates as observed at similar surface coverages on quartz. 5. D I S C U S S I O N

Comparison of the X-ray absorption spectra of Co sorbed to quartz and rntile at similar bulk surface coverages indi-

cates significant differences in the local atomic geometry around sorbed Co. For Co sorbed to quartz, the local structure is similar to that of Co in Co(OH)2(s) in which Co is octahedrally coordinated in a planar, closest-packed structure, although interatomic distances in the sorption samples are slightly contracted relative to those in Co(OH)2(s). The average number of second-neighbor Co atoms is ~6, the same as in Co (OH)2 ( s ), and this number is constant ( within error) for the range of surface coverages examined. Changes in Co concentration and pH of reacting solutions in the uptake experiments from undersaturated to oversaturated (with respect to all forms of Co (OH)2 ( s ) ) have no apparent effect on the local structure of the surface species. Likewise, equilibration for twenty-one days at relatively high pH produced no discernible change in the EXAFS compared to equilibration for 24 h (Tables 2, 3). These results indicate the formation of, primarily, large, multinuclear clusters of Co atoms or hydroxide-like precipitates in the presence of quartz which persist over a wide range of solution concentrations and for different equilibration times. The presence of Si secondneighbor backscattering from quartz in addition to Co is permitted in the analysis of the EXAFS spectra, but cannot be definitively proven. The strong backscattering from Co atoms masks backscattering from Si and precludes a detailed interpretation of the substrate structure relative to the sorbed Co species. The local atomic environment around Co sorbed to quartz is similar to that of Co sorbed to kaolinite (O'Day et al., 1994a) and alumina (Chisholm-Brause, 1991; Towle et al., 1995b) at high surface sorption densities. This environment, however, forms at lower surface coverage on quartz (F = 0.77 #mol m -2) than on alumina (1-" > 5 #mol m -2) or kaolinite (F > 3 #mol m-2). Transmission electron microscopy (TEM) of Co sorbed to c~-A12Osat high coverage suggests the presence of two populations of Co species, one adsorbed on alumina particles and one forming a separate phase (Towle et al., 1996). Similarly, the EXAFS results for Co sorbed to quartz might be explained by the formation of adsorbed mononuclear complexes and a separate Co phase on or near the quartz surface (discussed further below). In contrast, no separate phases were discernable with TEM in samples of Co sorbed to kaolinite at high surface coverages (O'Day, 1992), although such phases may not have been evident due to the similar plate-like morphology of kaolinite and Co(OH)z(S). The EXAFS spectra of Co on rutile at surface coverages similar to the lowest coverages obtained on quartz (F = 0.63 and 1.38 # m o l m 2) indicate no or few average Co secondneighbor atoms (Nco = 0 - 1 . 5 ) , but show strong backscattering from Ti atoms at two different distances. Changes in the EXAFS spectra with increasing surface coverage are due to increasing average numbers of Co second-neighbor atoms and changes in Co-Ti distances, but the Co/rutile spectra at high coverage do not resemble those of the quartz sorption samples nor Co(OH)z(S). These results imply the formation of primarily mononuclear and multinuclear complexes rather than large, hydroxide-like precipitates. For Co sorption on rutile, lower average numbers of second-neighbor Co backscatterers and stronger backscattering from Ti relative to Si allow an analysis of the structural relationship between sorbed Co and the rutile substrate. Differences in the EXAFS

Sorption of Co 2+ on quartz and rutile

2525

¢¢5,

(a)

-1C

3

4

5

6

7_1 8 k(/k )

9

11

10

12 R(,~)

¢.e5,

(b)

3

4

5

6

~-1

k(

8

9

10

11

)

12

R(A)

F1G. 5. (a) Normalized EXAFS spectra (solid line) and nonlinear least squares fit (dashed line) of Co/rutile sorption samples at surface coverages of 0.63 and 1.38 # m o l m -2. The fit models shown assumed first-neighbor O and two sets of second-neighbor Ti atoms. For the fit of the 1.38 #mol m -2 spectrum, a set of Co backscattering atoms, in addition to Ti atoms, must be included in the fit to match the Fourier transform spectrum (see Table 3 ). (b) Normalized EXAFS spectra (solid line) and nonlinear least squares fit (dashed line) of Co/rutile sorption sample at surface coverage of 4.65 #mol m -2. Bottom spectrum: Best fit assuming only first-neighbor O and second-neighbor Co backscattering atoms in the fit model. Top spectrum: Best fit assuming first-neighbor O and second-neighbor Co +

Ti

hnc'k~t'nttprlraer ntnmc

;n tho

~t mnttol

( ~

T~Kla

~

2516

P, A. O'D:avet :al.

FIG. 6. Possible geometries for a Co(OH)2(s)-like precipitate and mononuclear Co complexes with respect to the ideal quartz crystal structure. The EXAFS spectra indicate that most Co atoms are found in a local geometry similar to the planar, closest-packed structure of Co(OH)z(s) (right) that may or may not be attached directly to the quartz surface. A relatively small population of mononuclear or small multinuclear complexes may occupy defect sites or steps on the quartz surface where they are stabilized by multidentate bonds to Si tetrahedra (left).

spectra of Co sorbed to quartz and rutile can be interpreted by considering the polyhedral registry between octahedral Co sorption complexes and polyhedral elements of the bulk structure of the two substrates.

5.1. Sorption Sites on Quartz and Rutile

5.1.1. Quartz Previous solution and spectroscopic studies of metal ion uptake on quartz and amorphous silica have suggested that metal ions bond strongly to these surfaces, perhaps as innersphere complexes that form direct bonds between the metal ion and oxygen atoms of the Si tetrahedra. X-ray absorption studies of a number of metal cations sorbed to different oxide surfaces (summarized in Brown et al., 1995) indicate the prevalence of multidentate bonding rather than monodentate bonding between metal ions and oxide surfaces. Presumably, sorbed, charged ions are stabilized by a higher number of bonds to surface oxygen atoms, provided that the ion's coordination polyhedron is not significantly strained by forming bonds to the surface. Quartz is a framework silicate composed entirely of coruer-shared Si tetrahedra. Due to the large difference in O-O edge distances in Si tetrahedra (2.62-2.64 ,~) and in Co octahedra (O-O = 2.94-2.97 A), as well as metal-metal repulsion, it is unlikely that a Co octahedron bonded to the quartz surface would share a polyhedral edge with a Si tetrahedron, although this possibility is not precluded in principle as such polyhedral linkages are present in silicate minerals (e.g., olivine; Brown, 1982). A more likely geometry for direct attachment is comer-sharing between a Co octahedron and a Si tetrahedron. Furthermore, O-O distances between the comers of adjacent tetrahedra are all >3.3 A in the ideal structure of quartz, a distance too large for the coordination of vicinal oxygen atoms in a Co octahedron. Therefore, multidentate bonding of a Co octahedron would have to occur at either defect sites where Si tetrahedral corners are rotated to a closer distance, or at

merous microscopic and spectroscopic methods have demonstrated the roughness of mineral surfaces at the molecular level (e.g., Somorjai, 1981; Sunagawa, 1987; Hochella, 1990, 1995) and it is expected that defect sites and surface irregularities on quartz would provide the highest energy sites for the initial sorption of Co. However, only those defect sites which provide optimal O-O separations for polydentate bonding would be highly reactive for Co sorption and, presumably, this would be a small population compared with the total number of surface sites. Once these types of multidentate sites are filled, monodentate bonding to the surface might be expected to occur. Our EXAFS results, however, indicate the formation of large multinuclear complexes or a hydroxide-like precipitate, even at relatively low average surface sorption densities. Mononuclear complexes, if they existed as a significant population of sorbed Co, would be indicated by lower average numbers of second-neighbor Co atoms and by a stronger backscattering contribution from Si, especially at the lowest surface coverages. The EXAFS results imply that mononuclear complexes form only at very low surface sorption densities, at least below those that can be characterized with XAS using current synchrotron radiation sources. It is possible that these first-sorbed Co atoms provide nucleation sites for the formation and growth of larger multinuclear complexes. The lack of polyhedral registry between Co octahedra and Si tetrahedra of the quartz surface may favor bonding of Co to previously sorbed Co octahedra rather than to Si tetrahedra, which then leads to the formation of a hydroxide-like structure (Fig. 6). This lack of polyhedral registry between Co octahedra and Si tetrahedra in quartz may also explain why Co does not sorb to quartz until solution pH is well above the pHpznpc of quartz (Fig. 1). If Co sorption at surface coverages higher than only a few percent of monolayer is primarily due to hydrolysis and direct bonding to previously sorbed Co atoms, then sorption characteristics, such as the

Sorption of Co 2+ on quartz and mtile Co-hydroxide-like surface rather than solely by a hydrated and hydrolyzed quartz surface. 5.1.2. Rutile

The tetragonal structure of rutile is composed of alternating rows of edge- and corner-shared Ti octahedra (Fig. 7), in which all Ti sites and O sites in the structure are equivalent (Cromer and Herrington, 1955; Abrahams and Berstein, 1971 ). Each Ti atom is six-coordinated by O in a slightly distorted octahedron composed of four planar O atoms at 1.945 A and two apical O atoms at 1.985 A. Thus, each O is bonded to three Ti atoms, two as planar oxygen in adjacent edge-shared octahedra and one as the corner-shared apex of a third octahedron. In the ideal rutile structure, nearest secondneighbor Ti atoms are two at 2.96 A (edge-shared octahedra in the same row) and eight at 3.57 .~ (in adjacent rows linked through octahedral comers). The second-neighbor Co coordination indicated by EXAFS at the lowest Co/rutile surface coverages (Table 3) is consistent with Co occupying a site on the rutile surface corresponding to a Ti site in the bulk structure with Ti backscatters at 2.95-2.99 A and 3.60-3.63 A. The lower numbers of Ti backscatters around Co relative to that of Ti in rutile is evidence for Co remaining on the futile surface and not diffusing into the structure. The slightly longer distances between Co and Ti compared to Ti-Ti in rutile may result from the slightly larger size of the WCo2+ ion (ionic radius = 0.72 A) compared to WTi4+ (ionic radius = 0.68 ]~). There are two general structural positions for Co to occupy on the ideal rutile surface (Fig. 7): (1) Edge-sharing with Ti at the end of a chain of Ti octahedra (Col in Fig. 7) and (2) corner-sharing with the apices of two Ti octahedra (Co2 in Fig. 7). These two positions correspond to attachment on the (001) and (110) faces, respectively, although the same geometric relationships will hold on other crystallographic surfaces, as well as on steps and kink sites. The first position extends an existing "row" of Tit6 octahedra, while the second position initiates a new "row" in the rutile structure. A Co atom sorbed in the second position would have only two bonds to the substrate if there is not another row of adjacent Ti atoms to coordinate to the apical O atom of the Co octahedron (Fig. 7). Whether Co sorbs in either position with two, three, or four bonds to surface O atoms depends on the local structure of the surface, including its roughness, termination structure, and deviations from ideality. Once surface sites are occupied by Co with direct bonds to the rutile structure, these sites can act as new adsorption sites for Co to form a multinuclear complex. The secondneighbor Co-Co distance of 3.12 * derived from EXAFS analysis indicates edge-sharing of Co octahedra similar to Co bonding in the quartz sorption samples and in Co (OH)2 (s). For Co bonding in either of the two rutile sites described above, addition of another edge-shared Co atom in the same row would result in a Co-Co distance near 3.1 A, with deviations arising from distortions of the Co octahedron. The difference in charge and size between Co 2+ and Ti 4+, however, probably prevents extensive propagation of a Co-substituted futile structure which would eventually result in a large negative charge imbalance. This is in accord with the observation that significant substitution of foreign

2527

cations in rutile generally leads to a disordered rutile structure (summarized in Waychunas, 1991 ). With increasing Co surface coverage on rutile, the average number of second-neighbor, edge-shared Co octahedra increases and Co-Ti interatomic distances lengthen from 2.952.99 A to 3.01 A and from 3.60-3.63 A to 3.83 A. The two highest coverage Co/rutile samples ( F = 4.65 and 9.51 #m m 2), for which the longer Co-Ti distances are obtained, were equilibrated with Co aqueous solutions for eleven days. The increase in the average number of edge-shared Co octahedra indicates larger-sized multinuclear Co complexes. The longer Co-Ti distances observed in the sorption samples are similar to cation-cation distances in anatase, a Tit2 poolymorph, in which the shortest Ti-Ti distances are 3.04 A (N = 4) for edge-sharing Ti octahedra and 3.785 A (N = 4) for corner-sharing octahedra (Horn et al., 1972). The cation geometry in anatase differs from rutile in that Ti octahedra share four edges and four corners with adjacent Ti octahedra, rather than two edges (along a row) and eight comers (to adjacent rows) in rutile. The longer Co-Ti distances can be accounted for by sorbing Co to relatively unconstrained surface sites which allow for polyhedral relaxation, such as initiating a new row rather than bonding to the end of a row (i.e., Co2 and Co3 in Fig. 7). Locally, Ti and Co octahedra may rearrange to the geometry of anatase by edge-sharing, rather than corner-sharing, parallel to a row of Ti octahedra (110 face), which would give rise to the longer interatomic distances observed in anatase (Co3 in Fig. 7). The low average coordination numbers and high values of ~r~- obtained in the least-squares fit for Co-Ti backscatterers suggest a high degree of disorder in the average local atomic environment. This may indicate different populations or binding sites for sorbed Co with different interatomic distances to Ti, e.g., rutile-like geometries with shorter Co-Ti distances (Col and Co2) and anatase-like geometries with longer distances (Co3). With longer equilibration time, the local structure around sorbed Co on rutile at high coverages does not resemble the local structure of Co in Co (OH)2(s), i.e., it does not become more "hydroxide-like." These resuits are similar to the EXAFS results obtained for Co sorbed to kaolinite and equilibrated for up to forty-five days, during which time the local atomic structure around sorbed Co apparently became more disordered (see O'Day et al., 1994a).

5.2. Formation of Multinuclear Complexes on Quartz and Rutile The EXAFS results indicate clearly that large, hydroxidelike Co species form in the presence of powdered quartz and not in the presence of rutile under similar experimental conditions. A number of factors may contribute to this difference, including solution characteristics, properties of water near mineral surfaces, and surface structural considerations (discussed in O'Day et al., 1994a). There were some differences in solution composition during preparation of the quartz and rutile sorption samples that may have affected the EXAFS results. Some of the quartz EXAFS samples were equilibrated at slightly higher pH than the rutile samples (Table 2) because Co uptake occurs at higher pH on quartz than on rutile (at equivalent solution conditions; Fig. 1). For undersaturated EXAFS samples, the activity prod-

2528

P.A. O'Day et al.

FIG. 7. Possible sorption sites of Co atoms on the rutile surface. Cobalt may occupy a position on the surface that corresponds to a Ti site in the bulk structure, either at the end of a row of Ti atoms (Col), or by initiating a new row with the rutile structure (Co2). Cobalt may also bond to the surface through edge-sharing parallel to a Ti row (Co3). This latter geometric arrangement is similar to the bonding of Ti octahedra in anatase, a TiO2 polymorph. The relatively unconstrained positions of Co2 and Co3, relative to Col, may allow for greater relaxation of Co octahedra and account for the longer bond distances obtained in the EXAFS analysis.

ucts calculated from the compositions of the solutions after equilibration with the solid were below the solubility constants reported for stable or unstable forms of Co(OH)2(s) (Feitknecht and Schindler, 1963; Baes and Mesmer, 1976). Although care was taken in the preparation of the sorption samples to raise pH slowly and not inadvertently exceed Co(OH)2(s) saturation, we cannot entirely rule out local supersaturation and precipitation. If this occurred, the EXAFS results imply that the quartz surface stabilized a large, early-formed hydroxide-like species and that no changes in the EXAFS occurred after equilibration for 24 hours or 21 days, even in contact with solutions undersaturated with respect to C o ( O H ) : ( s ) . In the rutile sorption samples, in contrast, the EXAFS results show differences in the local Co environment among samples equilibrated for 24 hours and 11 days, suggesting that some reconstruction or local atomic rearrangement as a function of time may have occurred. Dissolution of SiO2 and precipitation of a metastable hydrous phyllosilicate have been suggested as one mechanism to explain the geometric relationships derived from EXAFS analysis between sorbed metal ions and silica substrates (Charlet and M anceau, 1994). If this occurred in our system, we might expect changes in the EXAFS spectra as a function of time and stronger evidence for Si backscattering, neither of which is observed. If Si is assumed to contribute to the EXAFS spectra, the Co-Si distances determined in this study (3.37-3.46 ]~) are substantially longer than the Ni-Si distances determined by EXAFS analysis for Ni sorbed to amorphous silica gel catalysts (3.27-3.31 A; Bonneviot et al., 1989a,b, 1990) and are also longer than typical Me(II)-

Si distances in clay minerals ( ~ 3 . 2 5 - 3 . 3 0 A for Me(II) = Mg 2÷, Fe 2÷, Ni2+; Brindley and Wan, 1975; Mellini, 1982; Mellini and Zanazzi, 1987). These longer distances are consistent with comer-sharing between Co octahedra and Si tetrahedra as described above, rather than linking of octahedral and tetrahedral structural sheets, typical of phyllosilicates, which produce shorter distances. These structural resuits and the lack of distinct backscattering from Si atoms are more consistent with the formation of two populations of sorbed Co, mononuclear or small multinuclear adsorbed complexes and hydroxide-like precipitates, rather than with the "neoformation" of hydrous silicates as the primary mode of sorption as proposed by Charlet and Manceau (1994). Structural difference among quartz, rutile, and a Co-hydroxide phase may partly account for differences in the local structure of sorbed Co. As discussed above, there are surface sites on rutile that can be occupied by Co which correspond to a continuation of the Ti sublattice. These may be surface sites that can accomodate differences in charge and size between Co 2÷ and Ti 4+ such as defects or steps. As more Co atoms populate the surface, they probably do not locally propagate the rutile structure due to charge imbalance and/ or size mismatch between Ti and Co octahedra. Instead, a more energetically favorable local structure is formed: either edge-sharing of Co octahedra as in Co(OH)2(s) or, as perhaps suggested by the data at high coverage, a local geometry more similar to that of Ti octahedra in anatase. In ruffle and anatase, Ti octahedra are both edge- and comer-shared and this geometry provides an abundance of potential sorption

Sorption of Co z+ on quartz and rutile sites for Co octahedra, which are only edge-shared in Co(OH)2(s). While there is an increase in the number of second-neighbor Co atoms with increasing coverage, the rutile spectra do not indicate large multinuclear complexes as seen on quartz, perhaps due in part to the abundance of favorable edge-shared sites on the rutile surface. In quartz, Si is only tetrahedrally coordinated by oxygen and there are no obvious structural sites that Co octahedra can occupy, leaving only defect sites as potential polydentate sorption sites and perhaps promoting multinuclear complex formation. As noted above, an unreconstructed quartz surface does not provide a good structural template for the planar structure of Co(OH)2(s). Thus we might expect formation of a separate phase anchored to the quartz surface at a few defect sites rather than extensive epitaxy. In the EXAFS analysis, we noted a slight contraction of Co-Co interatomic distances (3.11-3.12 ,~), relative to distances in Co(OH)2(s) (3.173 A ) . In addition to the results of this study, contraction of EXAFS interatomic distances relative to those in Co(OH)2(s) has been observed in the Co/kaolinite (O'Day et al., 1994a,b), Co/T-A1203 (Chisholm-Brause, 1991), and Co/c~-A1203 (Towle et al., 1995b) systems, but not in wet, homogeneously precipitated forms of Co(OH)2(s) (O'Day et al., 1994a,b). These results imply a relationship between the presence of a foreign substrate and the contraction of the local structure of the multinuclear Co sorbent; however, comparison of the quartz structure with the structures of the other substrates argues against extensive structural epitaxy as the only explanation for the observed contraction. Other possible explanations include edge effects from small particle size, a different metastable phase of Co(OH)2(s), or incorporation of small amounts of Si, molecular water, Na +, or NO3 (from the background electrotype) into the Co-hydroxide structure. We also note that the formation of large multinuclear complexes or a surface precipitate on quartz is not due to occupation of all possible proton-exchangable sorption sites by Co atoms. Site densities of Co on quartz for F = 0.77 and 1.6 #m m -2 samples are 0.46 and 0.96 Co atoms nm -2, respectively, assuming uniform distribution of Co atoms over the entire BET surface area. These values are substantially lower than site densities of ionizable protons determined by a variety of methods, which range from 4.2-11.4 sites nm 2 for quartz and silica gels (reviewed by James and Parks, 1982). This difference suggests that estimates of proton-ionizable sites do not necessarily correlate with densities of adsorption sites of metal surface complexes. 6. CONCLUDING REMARKS

From the EXAFS analysis of Co sorbed to powdered rutile and quartz, we suggest that Co occupies structural sites on the rutile surface that correspond to Ti sites in the bulk structure, i.e., Ti lattice-equivalent sites. This mode of sorption is more favorable for Co sorption on rutile than on quartz due to the similar cation size of Co 2+ and Ti 4÷ and the octahedral coordination of Ti in rutile. On defect-free quartz, there are no similarly favorable surface sites unless Co undergoes a coordination change to tetrahedral, which is unlikely for Co in aqueous solution (Magini et al., 1988) and for which there is no evidence in the Co/quartz EXAFS

2529

spectra. The spectroscopic results, however, also indicate disorder in Co-Ti backscattering correlations and suggest that, as Co atoms populate the surface, they do not propagate an ideal rutile structure, presumably due to charge imbalance and distortion of the CoO6 octahedra. For Co/rutile samples at higher surface coverages and equilibrated for eleven days, the EXAFS analyses indicate Co-Co correlations indicative of edge-shared Co octahedra and backscattering from Ti atoms at distances similar to those of Ti octahedra in anatase. We suspect that, with longer equilibration time at higher surface coverages, Co may have adopted a local geometry with Ti octahedra similar to that of the cation arrangement in anatase, which is a more open structure than ruffle and could more easily accommodate larger Co octahedra. Comparison of the EXAFS spectra of Co sorbed on quartz and rutile highlights how differences in crystal structure might influence local atomic geometry around sorbed species. Increasing evidence suggests that the formation of multinuclear metal complexes on oxide surfaces is common, but also that there are distinct differences in the size and geometry of these complexes, and whether or not surface precipitates form, among different mineral substrates (ChisholmBrause et al., 1990a,b; Roe et al., 1991; Charlet and Manceau, 1992; Manceau et al., 1992a; Fendorf et al., 1992; Waychunas et al., 1993; O'Day et al., 1994a,b; Spadini et al., 1994; Towle et al., 1995b). For Co sorption on rutile, our EXAFS results suggest a progression from mononuclear to multinuclear surface complexes with increasing total Co surface coverage. Similar progressions have been observed for Co sorption on alumina (Chisholm-Brause, 1991 ) and on kaolinite (O'Day et al., 1994b). In these systems, EXAFS spectra at very low surface coverages (<0.35 #m m -2) clearly show only mononuclear complexes. The EXAFS spectra of high coverage samples (i.e., F greater than ~ 3 /~m m -2) more closely resemble that of the Co/quartz spectra of this study and, in these spectra, backscattering from the substrate is masked by Co-Co backscattering (ChisholmBrause, 1991; O'Day et al., 1994b; Towle et al., 1995b). If a similar progression from mononuclear species to surface precipitate takes place on quartz, Co mononuclear complexes must form only at very low coverages and probably only at defect surface sites. The formation of large hydroxide-like Co species on quartz at the lowest surface coverages for which we could obtain EXAFS spectra may suggest that first-sorbed Co atoms provide nucleation sites for multinuclear surface complexes and that these sites are energetically preferable to Si-O monodentate sorption sites. Similarly, the possibility of some Si backscattering in the EXAFS signal (although difficult to detect) may be accounted for by a small population of Co atoms that are coordinated by Si tetrahedra, either at defect sites, from diffusion below the surface at cracks, or from a small amount of quartz dissolution and reprecipitation of a mixed Co/Si phase. The availability of favorable surface metal sites on rutile compared to quartz may also account to some extent for differences in the macroscopic sorption behavior between these two minerals, such as the lower pH of Co sorption on rutile despite its higher pHpnzp~. As charged metal ions sorb on surfaces, even at very low concentrations, they must influence the interfacial region, including the local structure of surface atoms of the substrate, sorbed water, and hydroxyl

2530

P.A. O'Day et al.

ions, and properties such as local charge or dielectric constant. This observation underscores the need to quantify energetic contributions from the local bonding e n v i r o n m e n t around the sorbate ion in the d e v e l o p m e n t of predictive sorption models. Adsorption of ions, either directly to substrate Me-O sites or to previously sorbed ions, involves changes in interfacial free energy that arise from conformational changes (e.g., local structural geometry, surface reconstruction, water and hydrogen b o n d i n g ) and electrostatic changes (e.g., adsorption of charged species, pH-dependent surface charging, dielectric properties); both effects involve atoms beyond those directly bound to the adion. In two formulations of surface complexation models, a Born solvation term dependent on a " s o l v e n t " dielectric constant in the interfacial region, either an estimate based on the dielectric constants of the bulk solid, the interface, and bulk water ( J a m e s and Healy, 1972c) or the dielectric constant of the solid only (Sverjensky, 1993), has been included to account for differences in sorption a m o n g different solid substrates, in addition to a standard Coulombic term. In the case of quartz and rutile with very different solid dielectric constants, the dielectric constant of the interfacial region between bulk solid and aqueous solution may vary considerably between those of the pure solid and pure solution, and would also change as adions are sorbed to the surface and form incipient solids. The general quantification of sorption processes on different mineral surfaces should evaluate the relative importance of conformational and electrostatic contributions and account for changes in structure and speciation of sorbed complexes with changes in solution composition, surface sorption density, and equilibration time. Acknowledgments--Parts of this study include work from the Ph.D. dissertations of PAO and CJC-B at Stanford University. We thank R. O. James (Eastman Kodak) and R. E. Johnson (Dupont) for supplying the batches of futile used in this study. R. Rea performed the cleaning of the quartz powder. We thank the staffs of SSRL and NSLS for their help with our XAS experiments, members of the Brown/Parks research group at Stanford University for assistance with data collection, and J. J. Rehr and S. I. Zabinsky, University of Washington, for their assistance with the application and interpretation of FEFF modeling in this study. We appreciate the helpful and insightful reviews of the manuscript by Kim Hayes and two anonymous reviewers. Financial support for this work was provided by the National Science Foundation through Grants EAR-8805440 and EAR-9105015, and by the Department of Energy through Grant DE-FG03-93-ER14347 (to GEB and GAP). Work was done in part at SSRL, which is operated by the Department of Energy, Office of Basic Energy Sciences. The NSLS is supported by the Department of Energy, Division of Materials Sciences, and Division of Chemical Sciences, under contract No. DE-AC02-76CH00016. Editorial handling: G. Sposito

REFERENCES Abrahams S. C. and Berstein J. L. ( 1971 ) Rutile: Normal probability plot analysis and accurate measurements of crystal structure. J. Chem. Phys. 55, 3206-3211. Anderson M. A., Palm-Gennen M. H., Renard P. N., Defosse C., and Rouxhet P. G. (1984) Chemical and XPS study of the adsorption of iron(II) onto porous silica. J. Colloid Interface Sci. 102, 328-336. Baes C. F., Jr. and Mesmer R. E. (1976) The Hydrolysis of Cations. Wiley. Basseni V., Burlamacchi L., and Martini G. (1979) Use of paramagnetic probes for the study of liquid adsorbed on porous supports.

Copper(ll) in water solution. J. Amer. Chem. Soc. 101, 54715477. Benjamin M. M. (1978) Effects of competing metals and complexing ligands on trace metal adsorption at the oxide/solution interface. Ph.D. dissertation, Stanford Univ. Benjamin M. M. and Leckie J. O. (1980) Adsorption of metals at oxide interfaces: effects of the concentrations of adsorbate and competing metals. In Contaminents and Sediments (ed. R. A. Baker), pp. 305-322. ACS. Bleam W. F. and McBride M. B. (1985) Cluster formation versus isolated-site adsorption. A study of Mn (II) and Mg (II) adsorption on boehmite and goethite. J. Colloid Interface Sci. 103, 124- 132. Bonneviot L. et al. (1989a) Investigation by EXAFS of the effect of pH on the structure of Ni 2+ ions impregnated on silica. Physica B 158, 43-44. Bonneviot L, Clause O., Che M., Manceau A., and Dexpert H. (1989b) EXAFS characterization of the adsorption sites of nickel ammine and ethylenediamine complexes on a silica surface. Catoly. Today 6, 39-46. Bonneviot L., Legendre O., Kermarec M., Olivier D., and Che M. (1990) Characterization by UV-vis-NIR reflectance spectroscopy of the exchange sites of nickel on silica. J. Colloid Interface Sci. 134, 534-547. Brown G. E., Jr. (1982) Olivines and silicate spinels. In Orthosilicates (ed. P. H. Ribbe): Rev. Mineral. 5 (2nd Edition), pp. 275381. Brown G. E., Jr., Calas G., Waychunas G. A., and Petiau J. (1988) X-ray absorption spectroscopy and its applications in mineralogy and geochemistry. In Speetroscopic Methods in Mineralogy and Geology (ed. F. C. Hawthorne): Rev. Mineral. 18, pp. 431-512. Brown G. E., Jr., Parks G. A., and O'Day P. A. (1995) Sorption at mineral/water interfaces: Macroscopic and microscopic perspectives. In Mineral Surfaces (ed. D. J. Vaughan and R. A. D. Pattrick), pp. 129-183. Chapman and Hall. Brindley G. W. and Wan H.-M. (1975) Compositions, structures and thermal behavior of nickel-containing minerals in the lizarditenepouite series. Amer. Mineral. 60, 863-871. Burlamacchi L. and Martini G. (1980) Molecular dynamics in water solutions adsorbed on porous solids studied by ESR of paramagnetic probes. In Magnetic Resonance in Colloid and Interface Science (ed. J. P. Fraissard and H. A. Resing), pp. 621-628. Reidel. Bye G. C., McEvoy M., and Malati M. A. (1982) Adsorption of copper (It) ions from aqueous solution by five silica samples. J. Chem. Tech. Biotechnol. 32, 781-789-2318. Bye G. C., McEvoy M., and Malati M. A. (1983) Adsorption of some divalent cations from aqueous solution on precipitated silica. J. Chem. Soc., Faraday Trans. 1 79, 2311-2318. Cases J. M., Doerler N., and Francois M. ( 1985 ) Influence de different types de broyage fin sur les mineraux: I. Cas du quartz. In X Veme Congres Intl. de Mineral. v. 1. Editions GEDIM., Cannes, France, 167-179. Charlet L. and Manceau A. (1992) X-ray absorption spectroscopic study of the sorption of Cr(II) at the oxide-water interface. If. Adsorption, coprecipitation, and surface precipitation on hydrous ferric oxide. J. Colloid Interface Sci. 148, 443-458. Charlet L. and Manceau A. (1994) Evidence for the neoformation of clays upon sorption of Co ( II ) and Ni ( II ) on silicates. Geochim. Cosmochim. Acta 58, 2577-2582. Chishohn-Brause C. J. ( 1991 ) Spectroscopic and equilibrium study of cobalt (tI) sorption complexes at oxide/water interfaces. Ph.D. dissertation, Stanford Univ. Chisholm-Brause C. J., O'Day P. A., Brown G. E., Jr., and Parks G. A. (1990a) Evidence for multinuclear metal-ion complexes at solid/water interfaces from X-ray absorption spectroscopy. Nature 348, 528 530. Chisholm-Brause C. J., Hayes K. F., Roe A. L., Brown G. E., Jr., Parks G. A., and Leckie J. O. (1990b) Spectroscopic investigation of Pb(II) complexes at the y-Al203/water interface. Geochim. Cosmochim. Acta 54, 1897-1909. Clark J. R., Appleman D. E., and Papike J. J. (1969) Crystal-chemical characterization of clinopyroxenes based on eight new structure refinements. Mineral. Soc. Amer. Spec. Pap. 2, 31-50.

Sorption of Co 2+ on quartz and rutile Cromer D. T. and Herrington K. (1955) The structures of anatase and ruffle. J. Amer. Chem. Soc. 77, 4708-4709. Dugger D. L., Stanton J. H., Irby B. N., McConnell B. L., Cummings W. W., and Maatman R. W. (1964) The exchange of twenty metal ions with the weakly acidic silanol group of silica gel. J. Phys. Chem. 68, 757-760. Farley K. J., Dzombak D. A., and Morel F. M. M. (1985) A surface precipitation model for the sorption of cations on metal oxides. J. Colloid Interface Sci. 106, 226-242. Feitknecht W. and Schindler P. (1963) Solubility constants of metal ioxides, metal hydroxides and metal hydroxide salts in aqeous solution. Pure Applied Chem. 6, 130-199. Fendorf S. E., Sparks D. L., Fendorf M., and Gronsky R. (1992) Surface precipitation reactions on oxide surface. J. Colloid Interface Sci. 148, 295-298. Fokkink L. G. J., De Keizer A., and Lyklema J. (1989) Temperature dependence of the electrical double layer on oxides: Ruffle and hematite. J. Colloid Interface Sci. 127, 116-131. Fuerstenau M. C., Elgillani D. A., and Miller J. D. (1970) Adsorption mechanisms in nonmetallic activation systems. Trans. Soc. Mining Eng., AIME 247, 11 - 14. Hadjiivanov K., Klissurski K., Kantcheva M., and Davydov A. (1991) State and localization of cobalt, nickel, and copper ions adsorbed on titania (anatase). J. Chem. Soc. Faraday Trans. 87, 907-911. Hashimoto K., Hiramoto M., Lever A. B. P., and Sakata T. (1988) Luminescence decay of ruthenium(II) complexes adsorbed on metal oxide powders in vacuo: Energy gap dependence of the electron-transfer rate. J. Phys. Chem. 92, 1016-1018. Hayes K. F. and Leckie J. O. (1987) Modeling ionic strength effects on cation adsorption at hydrous oxide/solution interfaces. J. Colloid Interface Sci. 115, 564-572. Healy T. W., James R. O., and Cooper R. (1968) The adsorption of aqueous Co(II) at the silica/water interface. In Adsorption from Solution (ed. W. J. Weberand Jr. and E. Matijevic); Adv. Chem. Ser. 79, pp. 62-73. Hochella M. F., Jr. (1990) Atomic structure, microtopography, composition, and reactivity of mineral surfaces. In Mineral-Water Interface Geochemistry (ed. M. F. Hochella, Jr. and A. F. White); Rev. Mineral. 23, pp. 431-512. Hochella M. F., Jr. (1995) Mineral surfaces: their characterization and their chemical, physical and reactive nature. In Mineral Surfaces (ed. D. J. Vaughan and R. A. D. Pattrick), pp. 17-60. Chapman and Hall. Horn M., Schwerdtfeger C. F., and Meagher E. P. (1972) Refinement of the structure of anatase at several temperatures. Zeits. Kristallogr. 136, 273-281. Hronsky V., Rakos M., Belak J., Juhar J., and Kazar D. (1978) EPR study of Mn 2+ ions adsorption on silica gel. Czech. J. Phys. B28, 1277-1286. Hronsky V., Rakos M., Belak J., Juhar J., and Kazar D. (1979) EPR study of Mn 2+ ions on silica gel. In Magnetic Resonance and Related Phenomena (ed. E. Kundla et al.), p. 310. Springer-Verlag. Hunter R. J. (1981) Zeta Potential in Colloid Science. Academic Press. James R. O. ( 1971 ) The adsorption of hydrolysable metal ions at the oxide-water interface. Ph.D. dissertation, Univ. Melbourne. James R. O. and Healy T. W. (1972a) Adsorption of hydrolyzable metal ions at the oxide/water interface. I. Co(II) adsorption on SiO2 and TiO2 as model systems. J. Colloid Interface Sci. 40, 42-52. James R. O. and Healy T. W. (1972b) Adsorption of hydrolyzable metal ions at the oxide/water interface. II. Charge reversal of SiO2 and TiO2 colloids by adsorbed Co(II), La(III), and Th(IV) as model systems. J. Colloid Interface Sci. 40, 53-64. James R. O. and Healy T. W. (1972c) Adsorption of hydrolyzable metal ions at the oxide/water interface. III. A thermodynamic model of adsorption. J. Colloid Interface Sci. 40, 65-81. James R. O. and MacNaughton M. G. (1977) The adsorption of aqueous heavy metals on inorganic minerals. Geochim. Cosmochim. Acta 41, 1549-1555. James R. O. and Parks G. A. (1982) Characterization of aqueous

2531

colloids by their electrical double-layer and intrinsic surface chemical properties. Surface and Colloid Sci. 12, 119-216. James R. O., Stiglich P. J., and Healy T. W. (1975) Analysis of models of adsorption of metal ions at oxide/water interfaces. Faraday Disc. Chem. Soc. 59, 142-156. James R. O., Stiglich P. J., and Healy T. W. (1981) The TiO2/ aqueous electrolyte system--applications of colloid models and model colloids. In Adsorption from Aqueous Solution (ed. P. H. Tewari), pp. 19-40. Plenum. Junta J. L. and Hochella M. F., Jr. (1994) Manganese (II) oxidation at mineral surfaces: A microscopic and spectroscopic study. Geochim. Cosmochim. Acta 58, 4985-4999. Katz L. E. and Hayes K. F. (1995a) Surface complexation modeling. I. Strategy for modeling monomer complex formation at moderate surface coverage. J. Colloid Interface Sci. 170, 477-490. Katz L. E. and Hayes K. F. (1995b) Surface complexation modeling. II. Strategy for modeling polymer and precipitation reactions at high surface coverage. J. Colloid Interface Sci. 170, 491-501. Kajiwara T., Hasimoto K., Kawai T., and Sakata T. (1982) Dynamics of luminescence from Ru (bpy)3C12 adsorbed on semiconductor surfaces. J. Phys. Chem. 86, 4516-4522. Lytle F. W., Sandstrom D. R., Marques E. C., Wong J., Spiro C. L., Huffman G. P., and Huggins F. E. (1984) Measurement of soft x-ray absorption spectra with a fluorescence ion chamber detector. Nucl. Instr. and Meth. 226, 542-548. Ludwig C. and Schindler P. W. (1995) Surface complexation on TiO2: I: Adsorption of H + and Cu 2÷ ions onto TiO2 (anatase). J. Colloid Interface Sci. 169, 283-289. Magini M., Licheri G., Paschina G., Piccaluga G., and Pinna G. ( 1988 ) X -ray Diffraction oflons in Aqueous Solutions." Hydration and Complex Formation. CRC. Malati M. A., McEvoy M., and Harvey C. R. ( 1982 ) The adsorption of cadmium(II) and silver(I) ions on SiO2 and on TiO2. Surface Tech. 17, 165-174. Manceau A., Charlet L., Boisset M. C., Didier B., and Spadini L. (1992a) Sorption and speciation of heavy metals on hydrous Fe and Mn oxides. From microscopic to macroscopic. Applied Clay Sci. 7, 201-223. Manceau A., Gorshkov A. I., and Drits V. A. (1992b) Structural chemistry of Mn, Fe, Co, and Ni in manganese hydrous oxides: Part I. Information from XANES spectroscopy. Amer. Mineral. 77, 1133-1143. Martini G. (1981) The state of water adsorbed on silica gel as determined by ESR of transition metal ion probe. J. Colloid Interface Sci. 80, 39-48. Matijevic E. and White D. W. (1984) Adsorption and desorption of hydrolysed metal ions. Colloids and Surfaces 9, 355-370. McBride M. B., Fraser A. R., and McHardy W. J. (1984) Cu 2÷ interaction with microcrystalline gibbsite. Evidence for oriented chemisorbed copper ions. Clays Clay Mineral. 32, 12-18. McMaster W. H., Del Grande N. K., Mallett J. H., and Hubbell J. H. (1969) Compilation of X-ray cross sections Ill; U.S. Atom. Energ. Comm. UCRL-50174. Mellini M. (1982) The crystal structure of lizardite IT: hydrogen bonds and polytypism. Amer. Mineral. 67, 587-598. Mellini M. and Zanazzi P. F. (1987) Crystal structures of lizardite1T and lizardite-2Hl from Coli, Italy. Amer. Mineral. 72, 943948. Mustre de Leon J., Rehr J. J., Zabinsky S. 1., and Albers R. C. ( 1991 ) Ab initio curved-wave x-ray-absorption fine structure. Phys. Rev. B 44, 4146-4156. O'Day P. A. (1992) Structure, bonding, and site preference of cobalt(II) sorption complexes on kaolinite and quartz. Ph.D. dissertation, Stanford Univ. O'Day P. A., Brown G. E., Jr., and Parks G. A. (1991) EXAFS study of aqueous Co(If) sorption complexes on kaolinite and quartz surfaces. In X-ray Absorption Fine Structure (ed. S. S. Hasnain), pp. 260-262. Ellis Horwood Ltd. O'Day P. A., Brown G. E., Jr., and Parks G. A. (1994a) X-ray absorption spectroscopy of cobalt(II) multinuclear surface complexes and surface precipitates on kaolinite. J. Colloid Interface Sci. 165, 269-289. O'Day P. A., Parks G. A., and Brown G. E., Jr. (1994b) Molecular structure and binding sites of cobalt(ll) surface complexes on

2532

P.A. O'Day et al.

kaolinite from X-ray absorption spectroscopy. Clays Clay Minerals 42, 337-355. O'Day P. A., Rehr J. J., Zabinsky S. I., and Brown G. E., Jr. (1994c) Extended x-ray absorption fine structure (EXAFS) analysis of disorder and multiple-scattering in complex crystalline solids. J. Amer. Chem. Soc. 116, 2938-2949. Ottaviani M. F., Ceresa E. M., and Visca M. (1985) Cation adsorption at the TiO~-water interface. J. Colloid Interface Sci. 108, 114-122. Parker R. A. ( 1961 ) Static dielectric constant of rutile (TiO2), 1.61060°K. Phys. Rev. 124, 1719-1722. Parks G. A. (1965) The isoelectric points of solid oxides, solid hydroxides and aqueous hydroxo complex systems. Chem. Rev. 65, 177-198. Parks G. A. ( 1975 ) Adsorption in the marine environment. In Marine Geochemistry (ed. J. P. Riley and G. Skirrow), pp. 241-308. Academic Press. Ragai J. and Selim S. I. (1987) Ion-exchange and surface properties of titania gels from Ti(III) solutions. J. Colloid Interface Sci. 115, 139-147. Rehr J. J. (1993) Recent developments in multiple-scattering calculations of XAFS and XANES. Jpn. J. Appl. Phys., suppl. 32-2, 8-12. Rehr J. J., Albers R. C., and Zabinsky S. I. (1992) High-order multiple-scattering calculations of x-ray-absorption fine structure. Phys. Rev. Lett. 69, 3397-3400. Roe A. L., Hayes K. F., Chisholm-Brause C. J., Brown G. E., Jr.. Parks G. A., and Leckie J. O. (1991) X-ray absorption study of lead complexes at c~-FeOOH/water interfaces. Langmuir 7, 367373. Schindler P. W., Ftirst B., Dick R., and Wolf P. U. (1976) Ligand properties of surface silanol groups. I. Surface complex formation with Fe 3+, C u 2+ , C O 2+ , and Pb 2+ . J. Colloid Interface Sci. 55, 469-475. Shannon R. D. (1993) Dielectric polarizabilities of ions in oxides and fluorides. J. Appl. Phys. 73, 348-366. Somorjai G. A. (1981) Chemistry in Two Dimensions: Surfaces. Cornell Univ. Press. Spadini L., Manceau A., Schindler P. W., and Charlet L. (1994) Structure and stability of Cd 2+ surface complexes on ferric oxide. J. Colloid Interface Sci. 168, 73-86. Sposito G. (1986) Distinguishing adsorption from surface precipitation. In Geochemical Processes at Mineral Surfaces (ed. J. A. Davis and K. F. Hayes); ACS Symp. Ser. 323, pp. 217-228. Stern E. A. (1988) Theory of EXAFS. In X -ray Absorption: Principles, Applications, Techniques of EXAFS, SEXAFS, and XANES (ed. D. C. Koningsberger and R. Prins); Chem. Anal. 9, pp. 2 3 51.

Sunagawa 1. (1987) Surface microtopography of crystal faces. In Morphology of Crystals (ed. I. Sunagawa), pp. 323-365. Terra. Sverjensky D. A. (1993) Physical surface-complexation models for sorption at the mineral-water interface. Nature 364, 776-780. Sverjensky D. A. (1994) Zero-point-of-charge prediction from crystal chemistry and solvation theory. Geochim. Cosmochim. Acta 58, 3123-3129. Teo B. K. (1986) EXAFS: Basic principles and data analysis. In Inorganic Chemist O, Concepts 9. Spfinger-Verlag. Tewari P. H. and Lee W. (1975) Adsorption of Co(ll) at the oxidewater interface. J. Colloid Interface Sci. 52, 77-88. Theis T. L. and Richter R. O. (1980) Adsorption reactions of nickel species at oxide surfaces. In Particulates in Water: Characterization, Fate, Effects, and Removal (ed. M. C. Kavanaugh and J. O. Leckie); Adv. Chem. Ser., pp. 73-96. ACS. Thompson H. A., Brown G. E., Jr., and Parks G. A. (1995) Low and ambient temperature XAFS study of U(VI) coordination in solids and aqueous solution. Physica B 208/209, 167-168. Towle S. N., Bargar J. R., Brown G. E., Jr., Parks G. A., and Barbee T. W., Jr. (1995a) Effect of surface structure on the adsorption of Co(II) on a-AI203: a glancing angle XAFS study. In Structure and Properties of Interfaces in Ceramics (ed. D. A. Bonnell et al.); Materials Research Soc. Syrup. Proc. 357, 23-29. Towle S. N., Bargar J. R., Persson P., Brown G. E., Jr., and Parks G. A. (1995b) XAFS study of Co(II) sorption at the a-A120,water interface. Physiea B 208/209, 439-440. Towle S. N., Bargar J. R., Brown G. E. Jr., and Parks G. A. (1996) Surface precipitation in the aqueous Co(II)/A1203 sorption system. J. Colloid Interface (submitted). Van Veen J. A. R. (1989) An inquiry into the surface chemistry of TiO2 (anatase). Zeitschr. Phys. Chem. Neue Folge 162, 215229. von Zelewsky A. and Bemtgen J. M. (1982) Formation of ternary copper(lI) complexes at the surface of silica gel as studied by EPR spectroscopy, lnorg. Chem. 21, 1771-1777. Waite T. D., Davis J. A., Payne T. E., Waychunas G. A., and Xu N. (1994) Uranium(Vl) adsorption to ferrihydrite: Application of a surface complexation model. Geochim. Cosmochim. Acta 58, 5465 -5478. Waychunas G. A. ( 1991 ) Crystal chemistry of oxides and hydroxides. In Oxide Minerals: Petrologic and Magnetic Significance (ed. D. H. Lindsley); Rev. Mineral. 25, pp. 11-68. Waychunas G. A., Rea B. A., Fuller C. C., and Davis J. A. (1993) Surface chemistry of ferrihydrite: Part 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta 57, 2251-2269. Yates D. E. (1975) The structure of the oxide/aqueous electrolyte interface. Ph.D. dissertation, Univ. Melbourne.