Journal of Colloid and Interface Science 294 (2006) 265272 www.elsevier.

com/locate/jcis

Mechanistic modeling of arsenic retention on natural red earth in simulated environmental systems
Meththika Vithanage a, , Rohana Chandrajith b , Athula Bandara c , Rohan Weerasooriya a
a Chemical Modeling Laboratory, Institute of Fundamental Studies, Kandy, Sri Lanka b Department of Geology, University of Peradeniya, Peradeniya, Sri Lanka c Department of Chemical Physics, Fritz-Haber Institute der Max Planck Gesellshaft Berlin, Berlin, Germany

Received 4 March 2005; accepted 18 July 2005 Available online 16 September 2005

Abstract Arsenic retention on natural red earth (hereafter NRE) was examined as a function of pH, ionic strength, and initial arsenic loading using both macroscopic and spectroscopic methods. Proton binding sites on NRE were characterized by potentiometric titrations yielding an average pHzpc around 8.5. Both As(III) and As(V)NRE surface congurations were postulated by vibration spectroscopy. Spectroscopically, it is shown that arsenite forms monodentate complexes whereas arsenate forms bidendate complexes with NRE. When 4 < pH < 8 and [total arsenic as As(III) or As(V)] = 0.385 mol/L both arsenite and arsenate exhibit near 100% adsorption for a 10-fold variation of ionic strength that is ascribed to inner-sphere complexation of surface bonding. Arsenite exhibits an apparent bond-switching mechanism from inner-sphere to outer-sphere at excess As(III) loading (total arsenic as As(III) or As(V)] = 38.5 mol/L. Competitive effect of arsenate for arsenite adsorption sites was observed when [initial As] = 0.385 mol/L. In dual adsorbate systems the As(III) was reduced over 20%, showing a competition of arsenite for arsenate binding sites (or vice versa). All experimental data were quantied with a 2pK generalized diffused layer model considering two site types for both protons and anions binding using reaction stoichiometries, as follows: FeOH + H+ FeOH+ 2, FeOH FeO + H+ , AlOH + H+ AlOH+ 2, AlOH AlO + H+ , FeOH(s) + H3 AsO3(aq) FeHAsO2 3(s) + H2 O, 4.744, 9.03, 7.229, 9.316, 6.798, 5.319, 11.88, 9.061.

AlOH(s) + H3 AsO3(aq) AlHAsO2 3(s) + H2 O, 2 FeOH(s) + H2 AsO 4(aq) FeHAsO4(s) + H2 O, AlOH(s) + H2 AsO4(aq) AlHAsO2 4(s) + H2 O,

2005 Published by Elsevier Inc. Keywords: Surface complexation; Generalized diffuse layer; Arsenite; Arsenate; Competitive adsorption; Natural red earth

1. Introduction Arsenic is a toxic, carcinogenic element and has a complex chemistry in the environment due to its contrasting oxidation states. Although it exists in four major oxidation states, As(V)
* Corresponding author. Fax: +94 0 81 2232131.

E-mail address: meththikav@yahoo.com (M. Vithanage). 0021-9797/$ see front matter 2005 Published by Elsevier Inc. doi:10.1016/j.jcis.2005.07.026

and As(III) species are predominant in natural aquatic environments [1]. Arsenic is found in both surface and subsurface water as a result of rock weathering, industrial waste discharge, volcanic emissions, fertilizer applications, mining, and smelting; however, it is more abundant in groundwater than in surface water. In oxic environments, arsenate [As(V)] is prevalent and 2 exists as a monovalent (H2 AsO 4 ) or divalent (HAsO4 ) anion. Arsenite [As(III)] exists as an uncharged molecule (H3 AsO3 )

266

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

and as an anionic (H3 AsO 3 ) species in a moderately reducing environment where anoxic conditions persist [2]. Although As(III) is unstable under aerobic environment, the transformation of As(III) to As(V) is kinetically controlled [3]. Therefore both arsenic species can occur in nature. Naturally occurring arsenic is commonly incorporated into a large number of minerals such as arsenopyrite, realgar, and orpiment. On the basis of geochemistry, four probable mechanisms have been proposed to describe the arsenic mobility in aquifers: (a) mobilization of arsenic due to direct oxidation of arsenopyrite minerals; (b) reductive remobilization of arseniccontaminated iron oxyhydroxides (FeOOH) in the subsurface as a result of enhanced draw down and recharge of aquifers; (c) leaching by bicarbonate ion; and (d) desorption of arsenic by phosphate that migrates into aquifers after fertilizer use [47]. Release of arsenic occurs rapidly once Fe and Mn dissolve under reducing environments or otherwise by desorption, which occurs after As(V) is reduced to As(III), because As(III) is not as tightly bound to the sediments [1]. According to recent studies arsenic concentration in surface and subsurface water is high, particularly in tropical regions, which include Argentina, Chile, southwestern Taiwan, northern Mexico, southern Thailand, Inner Mongolia, China, Vietnam, West Bengal in India, and Bangladesh [1]. Therefore removal of arsenic from drinking water is extremely important and had been given much consideration the world over [815]. Natural red earth of Sri Lanka (hereafter NRE) was selected for our research because it is a well-known fact that ironbearing minerals adsorb arsenic efciently [1,15,16]. The reactive binding sites of NRE were characterized with H-surface titrations. The adsorption of As(III) and As(V) on NRE were examined in batch mode as a function of pH, ionic strength and adsorbate concentration both in single and dual ion systems. The proton and arsenic adsorption data were quantied by a generalized diffuse layer model (DLM) after plausible arsenicNRE surface complexes were postulated with transmission FTIR spectroscopy. The surface-charging properties of the solidsolution interface have been examined in terms of surface complexation thereby assuming both 1-pK (surface >SOH, species +1/2 >SOH2 , >SO1/2 ) and 2-pK (species >SOH, >SOH+ 2, >SO ) approaches [2,813]. When compared to the 1-pK model, the 2-pK approach is widely used, particularly on metal (hydro) oxides. In this case the surface is hydroxalated in two consequent steps, >SOH + H+ >SOH+ 2 and >SOH >SO + H+ . The following factors further merit the selection of DLM for H+ and As(III) and As(V) adsorption data on NRE. (a) The single-plane modeling approach eliminates the estimation of several tting parameters associated with multilayer models without sacricing accuracy. (b) The surface chargepotential relationship is determined by GouyChapman theory. (c) The combination of high inner layer capacitance and low outer layer capacitance of multilayer models indicates that the two planes are embedded in a porous mineral surface layer of nite thickness, which is in agreement with a generalized single layer of representation incorporated in the DLM. (d) The generalized DLM is capable of handling several sur-

face sites simultaneously [17]. The DLM is both simpler and more versatile in that it corrects adsorption for ionic strength using GouyChapman theory [18], it does not demand an input capacitance, and it demands the smallest number of input parameters [19]. Hence, this provides an excellent platform for the calculation of interfacial properties of NRE and its adsorption behaviour. 2. Materials and methods 2.1. Natural red earth The NRE used for the study was obtained from the northwestern part (latitudes and longitudes of 8 14 50 N and 79 45 45 E) of Sri Lanka. NRE occurs as rounded and wellsorted quartz sand in a red clayey matrix with accessory ilmenite and magnetite. The brick red color indicates oxidizing conditions for the formation of red hematite. The NRE contains little Fe2+ (01%). However, Fe3+ content is typically high (>2.0%), sometimes reaching up to 6% [20]. XRD and XRF studies were carried out for identication of the mineralogical properties and the bulk chemical composition of NRE, respectively. Acids, bases, and stock solutions were prepared in deionized, distilled water. Precalibrated solutions of 0.103 M NaOH and 0.106 M HNO3 were used for pH adjustments. 5 M NaNO3 was used to regulate electrolytic concentration. Analytical grade sodium arsenite (Fluka, Switzerland) and sodium arsenate (Sigma, Australia) were used as sources for stock solutions. 2.2. Experimental procedure 2.2.1. Surface titration NRE was mechanically sieved prior to the experiments. Quantication of proton-binding sites was carried out by a conventional potentiometric titration method. A quantity of 20 g/L of <63 m fraction of NRE suspension was equilibrated well at the desired ionic strength for 24 h. Prior to the equilibration and throughout the titration the sample was purged with pure N2 (99.996%) to minimize CO2 contamination. Three titration experiments were performed on the basis of different electrolytic concentrations (0.1, 0.01, and 0.001 M NaNO3 ). The initial pH of NRE suspension was 6.0 and it was raised to 9 with 0.103 M NaOH before commencement of titrations. In order to minimize the solid dissolution the solution pH was kept above 4.0. The surface charge ( H ) was calculated using the equation [19] H = (CA CB + [OH ] [H+ ]) F , a S (1)

where H is the surface charge (C/m2 ), C A is added acid concentration, C B is added base concentration, [OH ] is the hydroxyl ion concentration, [H+ ] is the proton concentration, a is equilibrium OH and H+ ion concentration for a given quantity of solid used (g/L), F is the Faraday constant (96,500 C/mol), and S represents the specic surface area (m2 /g).

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

267

2.2.2. Adsorption edges A series of batch experiments were carried out to investigate adsorption characteristics of the NRE. The initial adsorption experiment was performed with varying NRE concentrations to obtain the optimal NRE concentration for maximum adsorption. NRE suspensions of varying solid concentration were spiked with arsenite or arsenate to yield a nal concentration of 0.077 mol/L. Based on these preliminary data we found that 5 g/L NRE showed optimal results, and hence were used in adsorption studies. Both arsenate and arsenite adsorption edges were constructed in NaNO3 according to the following method. A 5 g/L NRE suspension was prepared and equilibrated for 24 h after adjustment of the pH 9 in desired ionic strengths (0.1, 0.01, and 0.001) by 5 M NaNO3 . Then the batch solution was spiked with arsenite or arsenate to yield nal concentrations 0.385 and 38.5 mol/L. The system pH was decremented at 1.0 pH intervals up to 4. At each point, a 10-ml sample potion was transferred to a centrifuge tube. These tubes were equilibrated for 24 h at 100 rpm (EYELA B603 shaker) and the pH of the suspension was measured again and recorded. The solidsolution separation was carried out by membrane ltration (0.45 m pore size) for determination by hydride generator atomic absorption spectroscopy (Model GBC 933 AA Australia). 2.3. Transmission Fourier transform infrared spectroscopy Fourier transform infrared spectra (Model JASCO FTIR 410) of arsenicNRE systems were examined to determine the bonding behavior of single- and dual-ion arsenic on the NRE after treating them with KBr at a 10:1 ratio. 2.4. Data modeling Surface acidity and arsenic binding constants were determined by the FITEQL computer algorithm of Herbelin and Westall [21]. According to XRF data the NRE contains both Al and Fe-derived phases. These sites were active and designated as >FeOH and >AlOH respectively. The >SiOH, which corresponds to silica phase, was not considered since this phase was effectively covered with Al and Fe layers at varying degrees of thickness (the SEM data are not given). The experiment for the determination of cation exchange capacity (CEC) was carried out according to a procedure discussed by Anderson and Ingram [22]. The CEC of NRE was low (18 meq/100 g of soil). Hence arsenic retention by an exchange mechanism was not considered. The parameters used for data modeling are given in Table 1. 3. Results and discussion The XRD analysis conrmed that the crystalline silica phase is dominant in NRE (details were not shown). The XRF results (Table 2) showed that Al (as Al2 O3 ) and Fe (as Fe2 O3 ) are also present in signicant proportions, probably existing as an amorphous coating around silica grains, to the extent that the surface properties of silica are masked completely (SEM images are not

Table 1 Parameters used for optimization Parameter Surface area Equilibrium constants + H2 AsO 3 + H = H3 AsO3 Value 350 m2 /g 9.20 14.23 19.22 2.30 6.80 11.60 Reference Assumed [13] [13] [13] [14] [14] [14]

+ HAsO2 3 + H = H2 AsO3 3 2 AsO3 + H+ = HAsO3 + H2 AsO 4 + H = H3 AsO4 + = H AsO HAsO2 + H 2 4 4 2 + AsO3 4 + H = HAsO4

Table 2 Chemical composition of NRE Sample (%) SiO2 TiO2 Al2 O3 Fe2 O3 MnO MgO CaO K2 O P2 O5 RE 54.15 5.54 20.73 12.39 0.23 0.3 0.28 1.17 0.13 Sample (mg/L) Ba Ce Cr Ga Nb Ni Pb Rb Sr Th V Y Zr RE 315 826 128 21 81 76 37 58 69 151 345 105 4553

shown). Therefore only >AlOH and >FeOH sites, which are believed to abut from the surfaces of amorphous Fe or Al coatings, are treated as active. The >SiOH sites, which correspond to the hydroxylation of the silica phase, were not considered in the calculations, since we envisaged that these sites were not readily available for surface complexing. 3.1. Potentiometric titrations Fig. 1 shows the variation of surface charge of NRE, H suspensions as a function of pH in 0.1, 0.01, and 0.001 M NaNO3 . As shown in the gure, it is evident that the H is mainly controlled by the H+ and OH ions, which indicates that both surface hydroxyl functional groups of NRE, i.e., >AlOH and >FeOH, behave amphoterically. The pHZPC of NRE was estimated experimentally as 8.8 (Fig. 1). Either when pH < 4 or

Fig. 1. Variation of surface charge, , with pH based on three different ionic strengths. Solid suspension concentration was 5 g/L. Symbols represent experimental data; solid lines represent model calculations. Titration window ranged from pH 4 to 9. pHzpc = 8.5.

268

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

Table 3 Optimized FITEQL results obtained for potentiometric titration Parameter Site density FeOH AlOH Experimental pHZPC log K FeOH FeOH+ 2 FeO AlOH AlOH+ 2 AlO
AlOH(s) + H3 AsO3(aq) AlHAsO2 3(s) + H2 O FeOH(s) + H3 AsO3(aq) FeHAsO2 3(s) + H2 O

Value 0.1 M 1.309 102 6.600 104 8.8 0.01 M 1.136 103 4.217 104 0.001 M 3.042 102 1.160 103 Weighted meana 7.782 103 6.423 104

4.269 9.868 7.238 10.1 9.769 9.539 11.64 7.025

5.303 8.168 6.979 8.616 6.798 5.319 11.88 9.061

4.604 8.874 7.586 9.567 6.568 3.506 11.46 9.314

4.744 9.03 7.229 9.316

2 2FeOH(s) + 2H2 AsO 4(aq) 2FeHAsO4(s) + 2H2 O 2 2AlOH(s) + 2H2 AsO 4(aq) 2AlHAsO4(s) + 2H2 O a Weighted averages were calculated from log K = FITEQL calculations.

wi (log K)i , where wi = (1/log K )i /(

(1/log K )i , where log K is the standard deviation obtained from

when pH > 10, the solid shows some dissolution; hence reliable titration data cannot be obtained in these regions. The titration data obtained in the pH range 49 were, therefore, taken for calculations. Although operational, the generalized diffused layer model (DLM) assumed two types of surface sites with equal afnity for protons [17]. However, H+ -titration data of NRE cannot be successfully modeled by making such an assumption. Therefore, the proton afnity constants of >ALOH and >FeOH sites were assumed to be different. These constants and the site densities of >AlOH and >FeOH were obtained by numerical optimization using FITEQL [21]. The calculated data are shown in Table 3. The optimized ratio of >AlOH and >FeOH sites, i.e., >ALOH/>FeOH, always converged to 10. As shown in the data (Table 3), the intrinsic acidic constants of >AlOH sites were almost comparable to that of amorphous aluminum oxide (AAO), and those of >FeOH sites were similar to that of amorphous iron oxide (AFO). This is expected since the surface coating of NRE is thought of forming by entwining AAO or AFO layers around sand (silica) grains. The experimentally derived pHzpc value represents a composite effect of all active sites present on the NRE surface, i.e., >ALOH and >FeOH. The pHzpc was obtained as 8.5 by data modeling. The weights used were derived by optimization (Table 3). Moreover, the intrinsic acidity constants of different sites of NRE were averaged according to the procedure shown by Dzombak and Morel [17]. Table 3 shows nal results for the intrinsic acidity constants and site density values of NRE. 3.2. Adsorption of arsenic species onto NRE 3.2.1. Arsenite adsorption The adsorption density of arsenite on NRE, As(III) , reached almost 100% when the pH ranged between 4 and 8.0 at initial loading [As(III)] = 0.385 mol/L (Fig. 2A). When

pH > 8 the As(III) values showed around 70% reduction. As shown in Fig. 2B, for excessive initial loadings of As(III), i.e., 38.5 mol/L, the optimal As(III) values decreased by about 50%. The As(III) was optimal over the range pH 49 and no signicant decrease was observed with further increase of pH (up to 9). The AFO showed 100% As(III) adsorption for the same pH range at initial loading of [As(III)] = 20 mol/L [11]. In this case the solid suspension was 0.5 g/L, which corresponds to a site density value of 2.31 sites/nm. Parabolic behavior was shown for As(III) adsorption on AAO yielding a maximum around pH 8.5 when initial [As(III)] = 30 mol/L and solid content 1 g/L was used [11]. However, at an initial loading of As(III) of 100 mol/L and solid content of 4 g/L, the adsorption envelopes of both AAO and AFO were similar to that of NRE [2]. Based on these data it is concluded that the As(III) removal efciency was dependent on both initial loading and NRE content used. According to Hayes et al. [23], the mechanisms of surface complexation of anions (and cations as well) are postulated by examining the ionic strength dependence of the adsorption edges. According to these workers, outer spherically bonded surface complexes exhibit a marked effect on ionic strength, yielding distinctly separated adsorption edges. In the case of inner sphere bonding such a separation of adsorption edges cannot be readily discerned. However, these observations should be interpreted with caution. For example, how far should the adsorption edges be separated from a given point to yield 100% outer sphere or inner sphere complexes? Therefore, by visually judging the separation of adsorption edges, only the relative degree or dominance of inner- or outer sphere bonding can be inferred. Accordingly, the arsenite adsorption on NRE (at initial [0.385 mol/L]) does not show any ionic strength dependency, which evidences that an inner-sphere bonding mechanism is dominant (Fig. 2A). As shown in Fig. 2B, at high arsenite

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

269

Fig. 2. Variation of arsenite adsorption on NRE as a function of pH and ionic strength: (A) 0.385 mol/L; (B) 38.5 mol/L. Suspension density: 5 g/L. Solid, broken, and dotted lines represent DL model calculations where squares, circles, and triangles represent experimental data for 0.1, 0.01, and 0.001 M ionic strength, respectively.

Fig. 3. Arsenate adsorption on NRE as a function of pH and ionic strength: (A) 0.385 and (B) 38.5 mol/L. Suspension density: 5 g/L. Solid, broken, and dotted lines represent DL model calculations where squares, circles, and triangles represent experimental data for 0.1, 0.01, and 0.001 M ionic strength, respectively.

loadings well-separated adsorption edges were obtained, which signals that outer-sphere bonding is dominant. The relative degree of inner- or outer-sphere bonding of a given surface complex cannot, however, be forecast unambiguously from these data. Thus the stoichiometries that correspond to inner-sphere bonding of As(III) on NRE were taken into account in modeling. 3.2.2. Arsenate adsorption Arsenate adsorption on NRE was shown as nearly 100% for low surface loadings of arsenate (=0.385 mol/L) in the pH range between 4.0 and 8.0 (Fig. 3A). As(V) adsorption decreased to some extent (10%) with the additional increase of pH. Although the AsV values were reduced by about 30%, the trends of adsorption edges at high initial As(V) loadings (38.5 mol/L) were almost similar (Fig. 3B). The decrease of optimal AsV values at excessive As(V) loadings was ascribed to the availability of sites for adsorption. In contrast to As(III), As(V) adsorption edges show indifferent behavior for varia-

tions of ionic strength. Thus the As(V) on NRE is dominated by inner-sphere complexation at all adsorbate loadings examined. 3.3. FTIR results FTIR experiments were carried out at pH 4, 6, and 8 to observe the characteristics of NRE and its adsorption of arsenite and arsenate. FTIR studies were performed only for 0.01 M ionic strength. A signicant effect in the NRE spectrum is observed at pH 4, whereas the variations of spectra, when compared to those of bare NRE, were minimal at pH 6 and 8 (not shown). Attention was given to the region 40003000 cm1 of spectra since most changes are expected to observe in this region upon the sorption of both arsenite and arsenate. The spectrum of bare NRE showed a broad peak (Fig. 4A) with no characteristic peaks for terminal OH groups, which denotes absence of inner or outer terminal hydroxyl groups located at the surface of NRE at pH 4.

270

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

Peaks observed between 1250 and 650 cm1 are ascribed to metal oxide bonds such as >AlO and >FeO vibrations. Arsenite-treated NRE showed a characteristic peak in the region 40003000 cm1 (Fig. 4B) and the peak intensities are low in arsenate-treated NRE (Fig. 4C) samples. As(III)-treated NRE showed high-intensity peaks in the region 40003000 cm1 , which are ascribed to monodentate surface complexes. Deconvolution of the spectral data of arsenate resulted in seven distinct components (peak/width) at 3265/155, 3409/113, 3482/67, 3553/76, 3621/22, 3650/23, and 3697/16. However, the spectrum of arsenite treated NRE is reproduced with six peaks altogether, with the peak at 3650/23 disappearing. Myneni et al. [24] presented an arsenate reaction with freshly prepared hydrous iron oxide, which produced peaks at 700, 802, and 875, and for goethite 719, 730, 810, 838, and 938 cm1 . Only a few bands of these were observed in the arsenic-treated NRE (700900 cm1 ). The peak at 914 cm1 can be assigned as >AlOH vibration and it was intense in treated samples, which indicates >AlOAs bond formation. Peaks recorded in the region 600900 cm1 by Goldberg and Johnston [2] for Fe oxides did not appear in bare and treated NRE samples, but peaks in the region 1200900 cm1 were present, which can be assigned to surface complex formation. Sun and Doner [25] have shown that arsenate was adsorbed on goethite inner-spherically, forming strong bidentate binuclear complexes. This is proved by our sorption studies and the peak intensities observed at the 40003000 cm1 region in the FTIR studies. Therefore it can be postulated that the arsenate forms bidentate complexes on natural NRE. 3.4. Data modeling The modeling of both As(III) and As(V) adsorption onto NRE was carried out with the generalized diffuse-layer model with the modications stated in Section 3.1. In analogy to Dzombak and Morel [17], Bradbury and Baeyen [26] and Hoch and Weerasooriya [27], densities of >AlOH and >FeOH site ratio were kept around 1:100. The CEC of NRE was comparably low to that of swelling clays and hence not incorporated into the model. According to vibrational spectroscopic data it is postulated that As(III) forms monodentates, whereas As(V) form bidentate complexes with >AlOH and >FeOH sites on NRE. Ionic strength dependence data on adsorption support chemical bonding of both As(III) and As(V) on NRE. Collating all information, we introduced the following reaction stoichiometries to model both As(III) and As(V) adsorption data:
FeOH(s) + H3 AsO3(aq) FeHAsO2 3(s) + H2 O,

(2) (3) (4) (5)

Fig. 4. FTIR spectra of aqueous suspension of bare NRE and arsenic-treated NRE at pH 4 for the region 40003000 cm1 . (A) Bare NRE; (B) curve decomposition for arseniteNRE; (C) curve decomposition for arsenateNRE. Arsenite and arsenate were at equimolar concentration of 0.385 mol/L. Suspension concentration of 5 g/L for 0.01 M NaNO3 .

AlOH(s) + H3 AsO3(aq) AlHAsO2 3(s) + H2 O, 2 FeOH(s) + H2 AsO 4(aq) FeHAsO4(s) + H2 O, 2 AlOH(s) + H2 AsO 4(aq) AlHAsO4(s) + H2 O.

When compared to the bare spectrum of NRE, a marked difference of the peak intensities was observed in arsenite- and arsenate-treated NRE samples, particularly in the spectral regions 40003000, 17501250, 1100900, and 800650 cm1 .

The model calibrations were carried out using the experimental data set obtained for initial As(III) = As(V) = 0.385 mol/L. The mass action coefcients of reactions (2)(5) were obtained by optimization. The modeled data tted well with experimental values (solid, broken, and dotted lines in Figs. 2A

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

271

4. Summary Natural NRE has both >FeOH and >AlOH sites for surface complexation. The experimentally derived pHzpc is 8.5. Both arsenite and arsenate showed a strong afnity for NRE, having 100% adsorption between pH 4 and 7.5 at a solid concentration of 5 g/L. No pH or ionic strength dependency was observed in the pH range 47.5, suggesting an inner-sphere binding mechanism for low surface loads (0.385 mol/L). At excess adsorbate loads (38.5 mol/L), arsenite adsorption showed outer-sphere complex formation. Surface titration and adsorption data were modeled well with DLM. These ndings showed that both arsenite and arsenate species are efciently adsorbed on natural NRE surfaces and therefore natural NRE is a promising sorbent for developing an arsenic removal unit processes. Acknowledgments Professor Kazunari Domen (Tokyo University Japan) gave the JASCO FTIR spectra to Athula. We thank Professor N.W. Jern for providing SEM. Dharshani helped in the laboratories. Thanks go to Devika, Kushani Mahesh, and Madhu for their support given. Valuable comments made by anonymous reviewers signicantly improved the manuscript. References 3.5. Modeling competitive adsorption Further validations of the model were done using data generated from dual-sorbate systems. The competitive adsorption of anion species was obtained by spiking arsenite and arsenate into NRE suspension at an equimolar concentration of 0.385 mol/L. Nearly 100% adsorption was obtained for arsenate, while arsenite adsorption was decreased by about 5% in the pH range from 4 to 8. When pH > 8 both arsenite and arsenate showed decreases in adsorption (Fig. 5). In this pH range arsenite adsorption decreased up to 80% with the increase of pH and arsenate showed 40% decrease. However, the arsenic adsorption showed no distinct shifts with the ionic strength changes, which indicates dominance of inner-sphere complexes. The total site concentration of solids used is 1.58 103 mol/L; both arsenic species covered only onto 8.37 106 mol/L, showing that the NRE surface is largely undersaturated with respect to As(III) or As(V) adsorption. Under these circumstances the presence of As(III) affects the adsorption density of As(V); hence it can be assumed that As(V) and As(III) ions are competing for the same NRE sites. To model this competitive effect between As(V) and As(III) species, the same model was used to quantify the adsorption data using the mass action coefcients shown in Tables 1 and 3. As shown in Fig. 5 the competitive adsorption of As(III) and As(V) on NRE can be modeled well (solid, broken, and dotted lines) using single sorbate parameters obtained independently. Therefore it is concluded that this model is valid for competitive adsorption behavior of arsenic species as well.
[1] P.L. Smedley, D.G. Kinniburgh, Appl. Geochem. 17 (2002) 517. [2] S. Goldberg, C.T. Johnston, J. Colloid Interface Sci. 234 (2001) 204. [3] P.H. Masscheleyn, R.D. DeLaune, W.H. Patrick, Environ. Sci. Technol. 25 (1991) 1414. [4] R.T. Nickson, J.M. McArthur, P. Ravenscroft, W.G. Burgess, K.M. Ahmed, Appl. Geochem. 15 (2000) 403. [5] A.B. Mukherjee, P. Bhattacharya, Environ. Rev. 9 (2001) 189220. [6] M.H. Anwar, J. Akai, H. Sakugawa, Chemosphere 54 (2004) 753. [7] P. Bose, A. Sharma, Water Res. 36 (2002) 4916. [8] M.A. Anderson, J.F. Ferguson, J. Gavis, J. Colloid Interface Sci. 54 (1976) 391. [9] S. Goldberg, Soil Sci. Soc. Am. J. 50 (4) (1986) 1154. [10] A. Jain, R.H. Loeppert, J. Environ. Qual. 29 (2000) 1422. [11] S. Goldberg, Soil Sci. Soc. Am. J. 66 (2002) 413. [12] B.A. Manning, S. Goldberg, Soil Sci. Soc. Am. J. 60 (1996) 121. [13] R. Weerasooriya, H.J. Tobschall, D. Wijesekara, E.K.I.A.U.K. Arachige, K.A.S. Pathirathne, Chemosphere 51 (9) (2003) 1001. [14] O.S. Thirunavukkarasu, T. Viraraghavan, K.S. Subramanian, Water Qual. Res. J. Can. 36 (2001) 55. [15] H. Gen-Fuhrman, J.C. Tjell, D. McConchie, Environ. Sci. Technol. 38 (2004) 2428. [16] K.P. Raven, A. Jain, R.H. Loeppert, Environ. Sci. Technol. 32 (1998) 344. [17] D.A. Dzombak, F.M.M. Morel, Surface Complexation Modelling, Wiley, New York, 1990. [18] W. Stumm, J.J. Morgan, Aquatic Chemistry, WileyInterscience, New York, 1996. [19] D. Langmuir, Aqueous Environmental Geochemistry, PrenticeHall, Englewood Cliffs, NJ, 1997. [20] K. Dahanayake, S.K. Jayawardana, J. Geol. Soc. India 20 (1979) 433. [21] A.L. Herbelin, J.C. Westall, FITEQL: A Computer Program for Determination of Equilibrium Constants from Experimental Data, Report 99-

Fig. 5. Competitive adsorption of arsenic on NRE as a function of pH and ionic strength. Dual ion systems: equimolar concentration, As(III) = As(V) = 0.385 mol/L. Suspension concentration, 5 g/L. Solid, broken, and dotted lines represent DL model calculations; experimental data represented by symbols.

and 3A), and the optimized anion binding constants are reported in Table 3. The model validations were carried out using average mass action coefcients reported in Table 3 together with experimental data shown in Figs. 2B and 3B. As shown in Figs. 2B and 3B (solid, broken, and dotted lines), both As(V) and As(III) adsorption curves tted well for high adsorbate loading (38.5 mol/L).

272

M. Vithanage et al. / Journal of Colloid and Interface Science 294 (2006) 265272

01, Department of Chemistry, Oregon State University, Corvallis, OR, 1999. [22] J.M. Anderson, J.S.I. Ingram, Tropical Soil Biology and Fertility, CAB International, Wallingford, 1993. [23] K.F. Hayes, C. Papelis, J.O. Leckie, J. Colloid Interface Sci. 115 (1988) 564.

[24] S.C.B. Myneni, S.J. Traina, G.A. Waychunas, T.J. Logan, Geochim. Cosmochim. Acta 62 (19/20) (1998) 3285. [25] X. Sun, H. Doner, Soil Sci. Soc. Am. J. 161 (1996) 865. [26] M.B. Bradbury, B. Baeyen, Geochim. Cosmochim. Acta 66 (2002) 2325. [27] M. Hoch, R. Weerasooriya, Chemosphere 50 (2005) 743.