Journal of Colloid and Interface Science 273 (2004) 224233 www.elsevier.

com/locate/jcis

The titration of clay minerals I. Discontinuous backtitration technique combined with CEC measurements
Christophe Tournassat,a,b, Jean-Marc Greneche,c Delphine Tisserand,a and Laurent Charlet a
a LGIT-CNRS/UJF, University of Grenoble-I, P.O. Box 53, 38041 Grenoble, France b ANDRA, Parc de la Croix Blanche, 1/7, rue Jean Monnet, 92298 Chtenay-Malabry cedex, France c Laboratoire de Physique de LEtat Condens, UMR CNRS 6087, University du Maine, 72085 Le Mans cedex 9, France

Received 10 July 2003; accepted 7 November 2003

Abstract Previous experimental studies on clay potentiometric titration have been unable to distinguish inorganic cation exchange in the interlayer and on basal plane surfaces from specic pH-dependent sorption of cations and anions on the edges. In this study, we rened a titration technique, combining discontinuous backtitration and cation exchange capacity (CEC) measurements, and applied it to the potentiometric titration of Na- and Ca-conditioned montmorillonites. This technique can be used to accurately measure cation exchange, edge surface proton charge, dissolution of clay, and precipitation of new phases. Thus, a precise measurement of the variations of net proton surface charge is possible. This has important implications for clay surface modeling (see part II of this article) and for processes that depend on the clay surface charge, e.g., alteration, rheological processes, and contamination retention applications. In addition, this study conrms the adsorption of ionic pairs such as CaCl+ in exchange site positions and shows that CaOH+ could behave like CaCl+ . This result, together with the evidence of precipitation of a CaSi phase over a short time-scale (1 week) at high pH and low temperature, can be used to model clayconcrete interactions more accurately. We conrmed and quantied the H+ /Na+ exchange reaction at low pH. Finally, we demonstrate that both the edge surface charge and the permanent structural charge are compensated for by the nonspecic sorption of cations from solution across the entire pH range from 4 to 11. Under these conditions, the surface potential is fully screened and does not need to be invoked in modeling sorption processes on clay particles in dilute suspensions. 2003 Elsevier Inc. All rights reserved.
Keywords: Clay; Montmorillonite; Potentiometric titration; Back titration; CEC; Surface potential; Precipitation; Proton surface charge

1. Introduction Sorption mechanisms on clay surfaces have been studied for decades. One distinguishes between (i) inorganic cation exchange in the interlayer and on basal plane surfaces [1] and (ii) specic pH-dependent sorption of cations and anions on clay edges [2]. The edge sorption mechanism is a pH-dependent specic sorption on the clay edges. It has already been well established that the proton surface charge density (H ) on the clay edges depends on the physicochem* Corresponding author. Present address: BRGM, B.P. 6009, 3, avenue Claude Guillemin, 45060 Orlans cedex 2, France. E-mail address: c.tournassat@brgm.fr (C. Tournassat).

ical solution parameters (pH, ionic strength), which control the protonation state of the surface [310]. In contrast, cation exchange on basal planes is related to the permanent negative charge generated in the clay structure created by isomorphic substitutions in the lattice [1]. The density of exchange sites (0 ) can be derived from the structural formula, and is compensated for by exchangeable cations, which generally form outer sphere complexes with siloxane cavities [1,2]. In the literature, these two mechanisms are often analyzed separately. In this paper, we show the limitations of such experimental approaches, and then we present new experimental tools to study the two mechanisms side by side. This approach was used to constrain clay potentiometric titration data.

0021-9797/$ see front matter 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.jcis.2003.11.021

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

225

2. Experimental and theoretical background The rst attempts to titrate sorption sites on clay mineral edges were conducted by the same method as for oxides, continuous titration in a reactor [4,7,11]. The concentration of H+ bound to the surface is calculated according to the equation [7]: [H+ ] =
cK 1 w [H+ ]t [H+ ] + + C , a [H ]

(1)

where [H+ ] is the proton surface charge expressed in mol kg1 clay, a is the clay content of the suspension in kg clay l1 , [H+ ]t is the total amount of H+ in the system in mol l1 , [H+ ] is the concentration of protons measured in the suspension in mol l1 , c Kw is the dissociation constant of water as a function of temperature and ionic strength, and C is a correction factor used to consider side reactions like clay dissolution. In this approach, however, both the absolute value of [H+ ]t and C are difcult to dene, especially for clays. [H+ ]t must be considered as the sum [H+ ]t = [H+ ]init + [H+ ]added, where [H+ ]init is the initial unknown concentration of H+ present in the suspension at the start of the experiment, and [H+ ]added is the measured concentration of H+ added (or removed) during the titration. To calculate [H+ ], more than one titration curve must be recorded at different ionic strengths. In the case of oxides, titration curves recorded at different ionic strengths cross each other at a single point, the point of zero salt effect (PZSE), which is dened as equal to the point of zero net proton charge (PZNPC) ([12], and references therein). Hence, [H+ ]init can be calculated and the recalculated [H+ ] values can be considered as absolute values. In the case of clay minerals, however, titration curves recorded at different ionic strengths fail to yield a unique cross point [7,8,10]. Thus, [H+ ]init cannot be determined. This means that the [H+ ] values are only relative values [10,13], and strictly speaking, should be expressed as ( [H+ ]). The correction factor, C, when present, is not clearly described [7]. And since the dissolution rate of clay minerals as a function of pH is not well known, it is difcult to estimate the relevance of this correction factor. The increase in uncertainty linked to measurements at low and high pH values gives some additional difculties [7]. In previous studies, pH drifts below 1 mV h1 [4], 1 mV 5 min1 [7], and 1 mV min1 [11] were considered sufcient to assess the equilibrium of the protonation reaction. A value of 1 mV corresponds to 0.02 pH units, which is considered a standard error for pH measurements. Thus, according to Eq. (1), the error for [H+ ] due to the error for pH measurements is expressed as error = 1 Kw 1 + 2 pH (10pH 10pH+0.02 ). a 10 (2)

outside this pH range. At pH 2 or 12, the error value has the same order of magnitude as the total concentration of edge sites assumed to be present (100 mmol kg1 ) [4,9,14]. Finally, a fourth problem arises when the reversibility of the protonation reaction is tested. Baeyens and Bradbury observed a hysteresis phenomenon when they tried to titrate a clay suspension by increasing, and then decreasing, the pH [8,15]. This behavior was veried in a preliminary experiment in the present study. This hysteresis can originate from several processes including dissolution of the solid phase, or a carbonate effect, as pointed out by Gaboriaud and Ehrhardt in the case of a titration on goethite [16]. It could also originate from a kinetic effect. In continuous experiments with a short equilibration time, true equilibrium cannot be attained. An enhancement of H+ sorption can be measured by increasing the reaction time [7]. The backtitration technique avoids some of these problems. Schultess and Sparks [17] presented this technique for oxide surface titrations and Baeyens and Bradbury adapted it for clay surface titrations [8]. In this method, centrifugation tubes are lled with a known amount of clay suspension together with acid or alkali solutions. Once equilibrium has been reached (time scale depending on the system), and the pH has been measured, all samples are individually centrifuged and ltered. Each supernatant solution is weighed and the pH measured, and then the solution is backtitrated to pH 7, the titration endpoint. The difference between the amount of acid or base added at the start of the experiment and the amount of acid or base used in the backtitration yields the amount of H+ adsorbed or desorbed from the surface in eq kg1 , [H+ ] = 1 (VA CA VB CB ) a + (VBTA CBTA VBTB CBTB ) Vtot , Vback (3)

If a = 5103 kg l1 , the error is negligible between pH 4.5 to 9.5 (i.e., error <0.5 mmol kg1 ); however, this increases

where a = the clay content of the suspension in kg l1 ; Vtot = the whole volume of solution in one tube in l; Vback = the volume of ltrated suspension used in the backtitration in l; VA = the volume of acid solution added to the clay suspension in l; VB = the volume of alkali solution added to the clay suspension in l; CA = the concentration of acid solution in mol l1 ; CB = the concentration of alkali solution in mol l1 ; VBTA = the volume of acid solution used in backtitration in l; VBTB = the volume of alkali solution used in backtitration in l; CBTA = the concentration of acid solution used in backtitration in mol l1 ; CBTB = the concentration of alkali solution used in backtitration in mol l1 . The end point of the backtitration must be xed at pH 7 (or at a near neutral value) to prevent any dilution effect and to minimize the error in the volume of base or acid added. This technique, like the continuous titration technique, does not give an absolute position of the titration curve; hence, data obtained with this method [8] must be considered as relative titration curves as well [10,13].

226

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

Nevertheless, the backtitration technique minimizes many problems of the continuous titration technique. First, no kinetic effect is expected. Second, the analytical error is low even at low or high pH. Third, the effect of side reactions is minimized because the pH-dependent reactions of solute compounds and the dissolution of minerals are taken into account during the backtitration. But it is incorrect to assume that the dissolution of clay or other phases is counterbalanced by a reprecipitation of the same phases during the backtitration: Al oxides or amorphous silica are more likely to precipitate. Nevertheless, a signicant correction is obtained. Let us consider a congruent dissolution of the MX80 clay under acidic conditions (pH 4, structural formula from Sauzat et al. [18]): (Si7.96 Al0.04 )(Al3.10 Fe3+ Fe2+ Mg0.56 )O20 (OH)4 Na0.76 0.18 0.16 + 8.38H2O + 11.62H+ 7.96H4SiO0 + 3.14Al+3 + 0.16Fe2+ 4 + 0.18Fe(OH)3(s) + 0.56Mg2+ + 0.76Na+ . The backtitration to pH 7 leads to the formation of the following solute species: 7.96H4SiO0 7.96H4SiO0 , 4 4 3.14Al+3 + 9.42OH 3.14Al(OH)3(s), 0.16Fe2+ 0.16Fe2+ , 0.18Fe(OH)3(s) 0.18Fe(OH)3(s), 0.56Mg2+ 0.56Mg2+ , 0.76Na+ 0.76Na+ . Thus, 11.62H+ are consumed by the dissolution reaction and 9.42OH are consumed during the backtitration. The global H+ balance (+2.2H+ ) is the same in the case of alkali dissolution, (Si7.96 Al0.04 )(Al3.10 Fe3+ Fe2+ Mg0.56 )O20 (OH)4 Na0.76 0.18 0.16 + 12.04H2O + 5.92OH 7.96H3SiO + 3.14Al(OH)3(s) + 0.16FeOH+ 4 + 0.18Fe(OH)3(s) + 0.56Mg2+ + 0.76Na+ , with backtitration to pH 7 leading to the formation of the following solute species: 7.96H3SiO + 7.96H+ 7.96H4SiO0 4 4 or 7.96(SiO2(s) + 2H2 O), 3.14Al(OH)3(s) 3.14Al(OH)3(s), 0.16FeOH+ + 0.16H+ 0.16Fe2+ , 0.18Fe(OH)3(s) 0.18Fe(OH)3(s), 0.56Mg2+ 0.56Mg2+ , 0.76Na 0.76Na . are consumed by the dissolution reaction Here and 8.12H+ are consumed during the backtitration. Furthermore, the precipitation of Al(OH)3(s) or Fe(OH)3(s) does not 5.92OH
+ +

change the nal balance. The dissolution of silica phases is fully taken into account by this technique. Thus, the proton release or sink resulting from noncongruent dissolution is also taken into account by this method. Nevertheless, since trivalent cations generated by dissolution processes are partially removed from solution by cation exchange, they may not be backtitrated and the relative proton charge underestimated, as stated by Baeyens and Bradbury [8]. This effect can be minimized or avoided altogether by choosing a high Na+ or Ca2+ background salt concentration. Otherwise, a correction must be applied, or the pH domain, in which the dissolution occurs, cannot be analyzed. In this study, we rened the technique by combining the backtitration technique with cation exchange capacity (CEC) measurements, described by Sposito et al. [19,20]. The combination of the two methods was used to evaluate the relative proton charge variations together with the CEC variations as a function of pH. Protons sorbed onto the basal plane could be distinguished from protons sorbed onto clay edges, and the effect of proton uptake by clay edges on the overall CEC was evaluated.

3. Materials and methods 3.1. Chemicals All solutions and suspensions were prepared with boiled, argon-degassed Millipore Milli-Q 18 M water. NaOH and HCl stock solutions were made from Titrisol ampoules. NaCl and CaCl2 solutions were prepared from analytical grade salts. 3.2. Clay material preparation MX80 clay sample material (commercial Wyoming bentonite, reference BF100, CETCO France) was obtained from ANDRA (the French National Radioactive Waste Management Agency) after undergoing a homogenization treatment. The sample was dispersed in deionized water and the ne fraction (<2 m) isolated by sedimentation. After the suspension was saturated with NaCl (0.5 M), it was successively treated with a 0.1 M acetic acid, 0.5 M NaCl solution to remove carbonates, then with a dithionatecitratebicarbonate solution (DCB + 0.5 M NaCl), and nally with 3% H2 O2 , 0.5 M NaCl to remove mineral impurities and organic matter [21]. The nal suspensions were washed with either 0.5 M NaCl or 0.05 M CaCl2 solutions and then degassed with argon. 3.3. Clay material characterization 3.3.1. XRD data The raw material and the Na- and Ca-conditioned clay material were characterized with a Siemens D501 X-ray diffractometer equipped with a scintillation detector and

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

227

Fig. 2. Dots: 77 K Mssbauer spectrum of the Na-conditioned clay. Thin lines: decomposition of the spectrum. Thick line: sum of all the contributions of the decomposition. The parameters of the decomposition are given in Table 1. Contributions of Fe(III) A and Fe(III) B are shown together. The arrow indicates a strongly distorted base line at about 2 mm s1 . Fig. 1. Oriented XRD patterns of the ne fraction of the MX80 (full thin line), treated Na-MX80 (dotted line), and treated Ca-MX80 (dotteddashed line). The arrows indicate the remaining presence of cristobalite (4.03 ) and quartz (3.34 ) after treatment. Table 1 Mssbauer parameters of Na-conditioned amectite I.S. (mm s1 ) Fe(III) A Fe(III) B Singlet Fe(II) Magnetic sextet 0.45 (0.01) 0.45 (0.01) 0.45 (0.01) 1.26 (0.02) 0.45 (0.01) (mm s1 ) 0.47 (0.04) 0.47 (0.04) 4.2 0.27 (0.02) 1.65 Q.S. or 2 (mm s1 ) 0.50 (0.05) 1.12 (0.05) 0 3.07 (0.05) 0.14 (0.05) Bhf (T) % 20 (2) 10 (2) 41 (5) 11 (2) 18 (5)

diffracted-beam monochromated CoK radiation. Stepping rotor drive and data collection were performed with a Socabim DACO system. Oriented slides were prepared by pipetting a slurry of the clay suspensions onto a glass side and drying it at 40 C for a few hours to obtain an air-dried preparation. Results are shown in Fig. 1. The diffraction patterns of the conditioned material were found to be in agreement with Na- and Ca-saturated smectites and showed that few impurities (mostly quartz and cristobalite) were present after preparation. These impurities cannot be removed completely from natural clay, because they are as ne as the clay particles themselves, and they cannot be dissolved by chemical treatment without dissolving the clay particles. 3.3.2. Mssbauer data and chemical analysis The Mssbauer spectrum of the Na-conditioned smectite was recorded at 77 K using a constant acceleration spectrometer and a 57 Co source diffused into a rhodium matrix. Velocity calibrations were made using -Fe foil at 300 K. The hyperne parameters were rened using a least-squares tting procedure (unpublished program MOSFIT, Teillet and Varret, University of Maine, Le Mans, France). Results are shown in Fig. 2, and the associated hyperne parameters and error bars are listed in Table 1. In addition to the expected quadrupolar doublets of Fe(III) and Fe(II), the spectrum exhibits two unexpected features: a magnetic sextet with broadened lines, and a strongly distorted base line indicated at about 2 mm s1 by the arrow in Fig. 2. Thus, the tting model consists of three quadrupolar doublets (two attributed to Fe(III) sites and one to an Fe(II) site), and a single line and a magnetic sextet, both with broad lines. While other tting models could have been proposed, we concluded that the present one was the

50 (2)

Error bars are indicated in parentheses. I.S. = isomer shift; = linewidth; Q.S. = quadrupolar splitting value; 2 = quadrupolar shift value, Bhf = hyperne eld value, % = ratio of each component.

most reasonable. One criterion of goodness was the proportion of each iron type, taken as the absorption area of each component. The magnetic sextet is attributed to the presence of remaining iron oxides, while the single line is probably due to the presence of superparamagnetic Fe(III)-containing particles, resulting from an incomplete chemical reaction. The amount of oxides is too low to be characterized by XRD (18% of the total iron). Heron et al. [22] have demonstrated that the DCB treatment does not remove all the oxyhydroxides present in sediments, but that some large particles (e.g., goethite or hematite) are not completely dissolved. This result highlights the difculty of obtaining a pure smectite from a natural bentonite. In the absence of other information, the contribution of the singlet is attributed to Fe(III) present in the smectite structure. Based on this assumption, the clay structural Fe(II)/Fe(III) balance has been calculated as Fe(II)/Fetot = 11/82 = 13%. According to Sauzat et al. [18], the ne fraction of the MX80 montmorillonite has the following structural formula in the Na-saturated form, (Si3.98 Al0.02 )(Al1.55 Fe3+ Fe2+ Mg0.28 )O10 (OH)2 Na0.38 , 0.09 0.08 with the formula in the Ca-saturated expected to be (Si3.98 Al0.02 )(Al1.55 Fe3+ Fe2+ Mg0.28 )O10 (OH)2 Ca0.19 . 0.09 0.08

228

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

Table 2 Fe/Al and Mg/Al ratios (a) obtained by total digestion and chemical analysis, (b) before and after oxide content correction based on 77 K Mssbauer analysis, and (c) compared with Fe/Al and Mg/Al ratios obtained from the structural formula given by Sauzat et al. [18] Chemical analysis (a) Fe/Al molar ratio Mg/Al molar ratio 0.116 0.002 0.141 0.002 Mssbauer corrected chemical analysis (b) a 0.095 Structural formula (c) 0.108 0.025 0.178 0.032

a Based on the presence of 18% magnetic iron, attributed to residual iron oxides.

Table 3 Surface area values for the Na-MX80 clay fraction in suspension and in compacted state (after [23]) Basal surface area (m2 g1 ) Compacted clay Clay in dilute suspension 26.6 780a Interlayer surface area (m2 g1 ) 753 N.A.a Edge surface area (m2 g1 ) 8.5 8.5 Total surface area (m2 g1 ) 788 788

a The platelets are between one and two layers thick [23].

The Mssbauer results clearly show a discrepancy between the Fe(II)/Fe(III) balance of their clay and our sample. The pretreatment (described above) and the natural heterogeneity of the clay material may lead to this discrepancy. In particular, the DCB and the following H2 O2 treatments could have changed the Fe(II)/Fe(III) ratio compared to the original one. Thus, we decided to complete our clay characterization using a total digestion and chemical analysis by ICP-AES. The results of the chemical analysis are shown in Table 2. The measured Fetot /Al and Mg/Al ratios are compared to the ratio taken from the above-mentioned structural formula. The chemical results were corrected in agreement with Mssbauer results (Table 2): (i) 18% of the total iron, present as magnetic iron, was subtracted from the total iron to account for only the clay structural iron and (ii) the Fe(II)/Fetot balance was readjusted to 13%. By considering the same tetrahedral Al content, the structural formula was recalculated in order to be in agreement with the values given in Table 2. They are (Si3.98 Al0.02 ) (Al1.61 Fe3+ Fe2+ Mg0.24 )O10 (OH)2 Na0.28 for the Na-sat0.13 0.02 urated form, and (Si3.98 Al0.02 )(Al1.61 Fe3+ Fe2+ Mg0.24 ) 0.13 0.02 O10 (OH)2 Ca0.14 for the Ca-saturated form. 3.3.3. Surface characterization Various MX80 clay surface areas were characterized by Tournassat et al. [23] on the basis of derivative gas adsorption and AFM (Atomic Force Microscopy) experiments. The results are summarized in Table 3. 3.3.4. Clay content in suspension Titration results are most commonly given as a plot of surface proton density in mol kg1 vs pH (Eq. (3), [H+ ]), or as a plot of the surface charge, in C m2 , vs pH, H = F [H+ ], A (4)

modeled on the basis of structural interpretation. The measurement of the clay content in a suspension is not always easy. One of the classical methods is a weighing technique described by Sposito et al. [19]. In this method, an aliquot of the suspension is dried, cooled under a vacuum desiccator, and weighed. The measurement of salt concentration gives the amount of clay in suspension (= total weight salts weight). However, this technique has some limitations: some water is still attached to the clay platelets, even after drying (hydration water) [18], and worse, during the transfer from the desiccator to the balance, the clay can reabsorb water from the atmosphere. This can lead to an overestimate of the clay content in suspension. Here, we compare three methods of determining the amount of solid in our suspensions. The rst is described by Sposito et al. [19]. The second one is derived from a total chemical analysis by dissolution of the solid [24,25]. In this technique, a known volume of suspension containing approximately 100 mg of clay is put in a PtAu crucible. The suspension is dried at 110 C in an oven. Then, 750 mg of Li2 B4 O7 is added. The crucible is heated to 1000 C by an induction coil in a Philips Perl X-2 and then quickly cooled by pressurized air. The resulting glass is dissolved at 80 C in 20 ml of a 19% glycerin, 6% HNO3 solution and then diluted in 100 ml of water. The concentrations of Si, Al, Mg, Fe are measured by ICP-AES, and the amount of clay in the suspension is calculated based on the formula mclay = CAl Mclay , nAl (5)

where F = Faradays constant (96,480 C mol1 charge), and A = specic surface area. Both representations need a wellconstrained clay content value (see term a in Eq. (2)) to be

where mclay is the amount of clay in suspension (g l1 ), CAl the total measured concentration of Al in the suspension (mol l1 ), nAl the amount of Al in one clay structural formula unit, and Mclay the molar weight of one clay structural formula unit (733.3 g mol1 , based on the structural formula for our MX80 sample). In the third method, the clay suspension is diluted and then directly injected into an ICP-AES, which measures the

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

229

Al content. The second method (complete digestion) is used as the reference method. It was found that the third method underestimates the amount of solid by approximately 20%, depending on the dilution factor. We do not understand this result; hence, this method was not used for a precise estimate of the clay content. Nevertheless, the third method was used for another purpose described in Section 4. The rst method overestimates the amount of solid by about 18%. This difference is very close to the total amount of hydration water in the clay at room temperature (12% of the total mass, i.e., 14% of the dehydrated mass) [18]. Furthermore, the presence of silica impurities must be taken into account. The amount of solid in the suspension was determined either by the dissolution method or by the Sposito weighing method. In this latter case, 15% was subtracted from the total mass, in order to consider only the mass represented by the atoms present in the structural formula of the smectite. 3.4. Titration experiments Centrifugation tubes were numbered and precisely weighed (mtube in g) with a Mettler Toledo AG285 balance. A known volume of clay suspension was added to each 17ml centrifuge tube with a calibrated micropipette (Vsusp, clay content ). Volumes of NaOH and HCl were then added (VB and VA , concentration CB and CA ). Volumes of NaCl or CaCl2 solutions (VC , [H+ ] = CC ) were added until the total volume was 15 ml (Vtot ). The concentrations of NaCl and CaCl2 , and the clay contents are shown in Table 4 for each experiment. Care was taken to minimize the variations in salt ionic background concentrations from one tube to another in a given experiment. Nevertheless, these variations cannot be completely avoided because of cation exchange reactions between Na+ , Ca2+ , H+ , and ionic pairs (see Section 4). The tubes were shaken for 1 week. One experiment (0.05 M CaCl2 ionic background, experiment 3) was conducted in an O2 -free atmosphere glove box (Jacomex, pO2 < 0.5 ppm). The tubes were centrifuged after 1 week. An aliquot of supernatant uid from each tube was ltered (0.20 m) and used for backtitration (Vback ). The backtitrations were performed with a dosimeter (Metrohm 665 Dosimat, volume resolution = 0.005 ml), with pH 7 chosen as the titration end point. The pH was measured with a microelectrode (Mettler Toledo, inlab 423). Another aliquot was ltrated for Na, Ca, Si, Al, Mg, and Fe measurements on a PerkinElmer Optima 3300 DV inductively coupled plasma atomic emission spectrometer (ICP-AES). The Si,
Table 4 Experimental conditions of the three titrations Experiment 1 Ionic background Clay content NaCl 0.02 mol l1 4.69 g l1

Al, Mg, Fe concentrations of a third supernatant aliquot were measured without ltration to give the mass of suspended solids in the supernatant (mcorrection, obtained with Eq. (5); see second method in clay content determination section). The centrifuged reaction tubes, containing the clay slurry, were weighed (mcentrif ) and 10 ml of 1 M ammonium acetate was added to each tube. They were shaken for 1 week. Na- and Ca-CEC were then measured as described by Sposito et al. [20]. The relative proton surface charge [H+ ] is given by Eq. (3). The Na- and Ca-CEC, in eq kg1 , are given by Na-CEC =
Na Na 10 CAmm Csol (mcentrif mtube Vsusp) , Vsusp mcorrection (6) Ca Ca 10 CAmm Csol (mcentrif mtube Vsusp) , Vsusp mcorrection (7)

Ca-CEC = 2

where volumes (V ) are in ml, concentrations (C) in mol l1 , clay content in g l1 (), and masses (m) in g. The densities of all solutions were assumed to be equal to 1.0. Experimental conditions are summarized in Table 4.

4. Results and discussion Fig. 3 presents the results of the surface proton charge titrations with the CEC measurements. [OH ] is plotted as a function of pH to compare the variations of surface charge with those of CEC. [OH ] is the sum of + ] (Eq. (3)) plus an arbitrary constant [H c , chosen for a given experiment to match [OH ] and CEC values as closely as possible. As stated in the experimental and theoretical background section, the initial amount of sorbed H+ ([H+ ]init) at the start of the titration experiment is not known. Thus, [H+ ] values are only relative values [10,13]. The constant value c is different for experiments 1, 2 and 3, because the unknown initial protonation state is different for each experiment [13]. Only the relative and not the absolute value of [OH ] was measured, as the experiments cannot provide a true surface charge value in the absence of ZPSE. Error bars were calculated based on a standard error of 0.05 ml for volumes VBTA and VBTB , a standard error of 1% on other volumes, and a standard error of 2% on ICP-AES concentration measurements. Due to the magnitude of error in the CEC measurements, it was not possible to infer a relationship between

Experiment 2 CaCl2 0.0068 mol l1 3.03 g l1

Experiment 3 CaCl2 0.05 mol l1 1.53 g l1

230

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

Fig. 4. Relationship between the variation of proton surface charge (presented as [OH ]) and the variation of CEC ( CEC). The reference point ( CEC = OH = 0) is at pH 7.3. (!) experiment 1. (P) experiment 2. Experiment 3 is not represented because the associated error bars are too large to infer any relationship. The dotted line represents a 1:1 relationship.

Fig. 3. Titration measurements ( [OH ] (F)) presented concomitantly with the CEC measurements (1) as a function of pH. (1) Experiment 1, 0.02 mol l1 NaCl; (2) experiment 2, 0.0068 mol l1 CaCl2 ; (3) experiment 3, 0.05 mol l1 CaCl2 .

the constant c and 0 . We can only note that the constant value c is slightly higher than the structural CEC value for experiment 1 (constant = 0.82 eq kg1 , structural CEC = 0.76 eq kg1 ). Both experiments 1 and 2 exhibit experimental points at pH value 7.3. This pH value was used as a reference point to build Fig. 4, and we set [OH ] = CEC = 0 at this

point. [OH ] and CEC are well correlated: the straight line represents a one-to-one relationship and the experimental points lie on this line. This means that a decrease in the surface proton charge is counterbalanced by an equivalent increase in the Na+ or Ca2+ charge. Thus, protonated and deprotonated sites are not present as free charged species, but rather they are neutralized by the ions present in the neighborhood of the surface. This behavior was already suggested by Charlet et al. [4] on the basis of K-saturated montmorillonite continuous titration, where proton charge was almost independent of ionic strength. According to these authors, this result agrees with electroacoustic measurements performed on kaolinite, indicating that most of the surface charge is balanced by monovalent ions [26]. Hence, the surface complexes contribute nothing to the electrokinetic charge. This could explain why Bradbury and Baeyens needed to suppress the electrostatic term in order to t their H+ sorption curves concomitantly with their Zn2+ , Ni2+ , and Eu3+ sorption curves [9,27,28].

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

231

Fig. 5. Variation of the proton surface charge (presented as [OH ]) as a function of pH for experiments 1 (!), 2 (P), 3 (E), and the 7 days experiments of Baeyens and Bradbury ([8], "). The reference point ( [OH ] 0) is at pH 7.3.

Fig. 6. Concentration of Si in solution as a function of pH for experiments 1 (!) and 2 (P). The arrow indicates the decrease of the Si concentration at high pH in experiment 2 (CaCl2 ionic background) and the probable precipitation of a CaSi hydrated phase.

Fig. 5 shows the titration data of experiment 1 (NaCl 0.02 M), together with the 7 days, 0.5 M NaClO4 titration curve of Baeyens and Bradbury [8,15]. Baeyens and Bradburys experiments were conducted on Swy1 montmorillonite, the surface properties of which are not necessarily identical to those of the MX80. Nevertheless, there is little difference between the two titration curves in the pH range 49. Hence, we can consider that our results can be applied to their system and that the background anion (Cl vs ClO ) 4 has little impact on the titration data in this pH range. Fig. 5 also shows the titration data of experiments 1, 2, and 3 from pH 4 to 9 on the same graph. Between pH 4 and 9, there is no signicant difference between the results of the two Ca-clay titration data, and between the results of Na- and Ca-clay titration data. This result agrees with the previous statement: the ionic strength and composition of the background salt has no effect on the titration curve in the pH range 4 to 9. However, some pronounced differences among experiments appear below pH 4 and above pH 9 (Fig. 3). Cation exchange and dissolution most likely explain these differences. At pH > 9 and with Ca2+ as the cationic background, the large increase in [OH ] is correlated to the same increase in apparent Ca-CEC (Fig. 3). This increase is particularly pronounced between pH 11 and 12, i.e., around the rst hydrolysis constant of Ca2+ (pKa = 12.85, Llnl.dat database [29]). Thus, it is likely due to the sorption of CaOH+ cations. In the calculation of the Ca-CEC (Eq. (7)), each Ca2+ ion is assumed to balance two negative charges of the exchanger. Instead, the reaction Ca2+ + H2 O CaOH+ + H+ leads to (i) a net consumption of one OH per CaOH+ in exchange position and thus, to the increase of OH , and (ii) to an overestimation of the CEC, as each Ca balances only one charge of the exchanger. Nevertheless, the amount of CaOH+ in exchange position cannot exceed the CEC value. In the extreme case, where CaOH+ occupies all the exchange sites, the apparent CEC value should be twice

Fig. 7. Concentrations of Al (2), Fe (P), and Mg (!) as a function of pH in experiment 3.

the structural CEC value (1.52 eq kg1 based on the structural formula). In experiment 2, the amount of sorbed Ca2+ is equal to 1.85 eq kg1 at pH 11.93. Thus, an additional mechanism must be invoked, to explain this excess of apparent surface charge. Since this excess of charge is observed in a pH range, where a decrease in aqueous Si is observed (Fig. 6), the additional mechanism is attributed to the precipitation of a CaSi hydrate solid phase, e.g., a tobermoritelike phase (Ca5 Si6 H6 O20 ) [30]. The ion activity product observed in the experimental conditions of the last point of experiment 2 is equal to 1072.46, and indicates a net supersaturation with respect to tobermorite (Ca5 Si6 H6 O20 + 10H+ 5Ca2+ + 6SiO2 + 8H2 O pKs = 69.08). At pH < 4, the pronounced decrease in [OH ] concomitant to a Na-CEC decrease (Fig. 3) can be attributed to two distinct phenomena. The rst one is an H+ /Na+ exchange reaction [31]. The second one is the dissolution of the clay and the exchange of Na+ for Al3+ [8]. Fig. 7 shows the concentrations of Al, Mg, and Fe between pH 2 and 4.5

232

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

Fig. 8. Concentration of Al as a function of pH in experiments 1 (!), 2 (P), and 3 (E).

Fig. 9. Comparison of the CEC as a function of pH for experiments 1 (!), 2 (P), and 3 (E).

in experiment 3. There is a signicant release of Al to solution in this experiment (up to 2.9 105 mol l1 at pH 2.14). The Al concentrations in experiments 1, 2, and 3 are shown in Fig. 8. It appears that signicant and equivalent concentrations of Al are measured in the case of experiments 2 and 3, i.e., experiments in a CaCl2 ionic background. In the case of experiment 1, i.e., in an NaCl ionic background, the observed Al concentrations are close to the detection limit (0.5 mol l1 ). These low Al concentrations are attributed to the adsorption of Al3+ at the basal plane cation exchange position, due to the high Al/Na selectivity coefcient of montmorillonite (see [4], and the tabulation of selectivity coefcients in the Llnl.dat, Wateq4f, or Phreeqc databases [29]; log K Al/Na values range from 0.41 to 0.67). As a rst approximation, we consider that the selectivity coefcient of the Al/Ca exchange can be calculated by combining the Al/Na selectivity coefcient with the Na/Ca selectivity coefcient. This approximation is valid if the exchange processes on clay minerals can be considered as ideal processes. In the 1980s, Sposito et al. [1,3,19,20] showed the ideal behavior of Na+ divalent cation exchange reactions. This ideality is assumed here for a Na+ Al3+ exchange at low pH. Since the selectivity of the Ca/Na exchange is high (log K Ca/Na = 0.8 after Llnl.dat, Wateq4f, and Phreeqc databases), the selectivity coefcient for the Al/Ca exchange must be low (log K Al/Ca value from 1.58 to 1.06 according to the combination of the above cited values). Then, Al3+ will not displace Ca2+ from the exchange sites. This phenomenon explains the differences of Al3+ concentrations measured in the acidic pH range, between experiment 1 and, experiments 2 and 3. It is therefore possible to correct the apparent CEC and [OH ] values from this effect. We assume that the amount of Al3+ released in solution by dissolution is equal to the amount released in experiments 2 and 3 at the equivalent pH. The CEC and [OH ] values of experiments 1 (0.02 mol l1 NaCl) are then corrected between pH 1 and 4 by the difference of aqueous Al3+ moles of charge in solution (mol kg1 ) between experiment 1 and experiment 2 or 3

(0.0068 and 0.05 mol l1 CaCl2 , respectively). The correction does not exceed 0.02 mol kg1 . Despite this correction, the apparent CEC differences remain signicant between experiment 1 and experiments 2 and 3 (Fig. 9). Results of the [OH ] and apparent CEC measured in experiments 1, 2, and 3 indicate that H+ may replace Na+ in experiment 1, whereas the H+ Ca2+ exchange reaction is very limited in experiment 2 and 3. We also noted that the CaCEC is even greater than the Na-CEC, even within the pH range 47, where neither H+ nor CaOH+ sorption occur at the exchange site position (Fig. 9). This difference between Na-CEC and Ca-CEC has been ascribed to the adsorption of CaCl+ ionic pairs [20]. Previous studies on exchange reactions involving Na+ , + , and Cs+ demonstrate that a clay minerals CEC varies K as a function of pH [3,31,32]. In this article, we demonstrate that these variations are correlated with protonation/deprotonation reactions attributed to the reactivity of clay platelet edges, at pH values between 4 and 9. The CEC is usually measured at near neutral pH [33]. Under such conditions, the measured CEC must be attributed to the contribution of the structural charge ( O ) given by the structural formula and the contribution of the edge charge (H ).

5. Conclusions The combination of titration plus backtitration (with ammonium acetate extraction for Na- and Ca-CEC measurements) is a powerful technique for measuring the net proton surface charge excess. It identies the main mechanisms responsible for changes in titration curves: dissolution processes, variations in edge net proton surface charge, exchange mechanisms, adsorption of ionic pairs, and precipitation of new phases. To summarize, we have shown that between pH 4 and 9, titration data depend on neither the ionic background salt (NaCl or CaCl2 ) nor its concentration. Below pH 4, dissolu-

C. Tournassat et al. / Journal of Colloid and Interface Science 273 (2004) 224233

233

tion and Na+ /H+ exchange affect the titration data. Above pH 9, sorption of CaOH+ ionic pairs in the exchange site position can lead to overestimation of both the amounts of Ca2+ adsorbed or H+ desorbed (i.e., an overestimate of both the CEC and the net proton surface charge). Above pH 11, the dissolution of clay together with the presence of calcium in solution leads to the formation of a tobermorite-like phase. And across the entire pH range, 411, the negative charge created by the deprotonation reactions is compensated for by equivalent sorption of Na+ or Ca2+ . Hence, the electrostatic term can be ignored in clay complexation models. Our results were used to test two existing models and to construct a coherent H+ sorption model. This work is presented in Part II of this article.

[9] [10] [11] [12]

[13] [14] [15]

[16] [17] [18]

[19]

Acknowledgments
[20]

This research was funded by the French National radioactive waste management agency (ANDRA). French Geological Survey (BRGM) is acknowledged for its editorial nancial support.

[21] [22] [23]

References
[1] G. Sposito, The Thermodynamics of Soil Solution, Oxford Univ. Press, New York, 1981. [2] G. Sposito, Surface Chemistry of Soils, Oxford Univ. Press, New York, 1984. [3] P. Fletcher, G. Sposito, Clay Miner. 24 (1989) 375. [4] L. Charlet, P.W. Schindler, L. Spadini, G. Furrer, M. Zysset, Aquat. Sci. 55 (1993) 1015. [5] M. Stadler, P.W. Schindler, Clays Clay Miner. 41 (1993) 288. [6] J.M. Zachara, S.C. Smith, J.P. McKinley, C.T. Resch, Soil Sci. Soc. Am. J. 57 (1993) 1491. [7] H. Wanner, Y. Albinson, O. Karnland, E. Wieland, P. Wersin, L. Charlet, Radiochim. Acta 66/67 (1994) 157. [8] B. Baeyens, M.H. Bradbury, J. Contam. Hydrol. 27 (1997) 199.

[24] [25] [26] [27] [28] [29]

[30] [31] [32] [33]

M.H. Bradbury, B. Baeyens, J. Contam. Hydrol. 27 (1997) 223. M.J. Avena, Encyclopedia of Surface and Colloid Science, 2002, p. 37. M.J. Avena, C. De Pauli, J. Colloid Interface Sci. 202 (1998) 195. J.P. Jolivet, De la Solution lOxyde. Condensation des Cations en Solution Acqueuse. Chimie de Surface des Oxydes, InterEditions/CNRS Editions, Paris, 1994. G. Sposito, Environ. Sci. Technol. 32 (1998) 2815. J.M. Zachara, S.C. Smith, Soil Sci. Soc. Am. J. 58 (1994) 762. B. Baeyens, M.H. Bradbury, A Quantitative Mechanistic Description of Ni, Zn and Ca Sorption on Na-Montmorillonite, Paul Scherrer Institut (PSI), 1995. F. Gaboriaud, J.J. Ehrhardt, Geochim. Cosmochim. Acta 67 (2003) 967. C.P. Schultess, D.L. Sparks, Soil Sci. Soc. Amer. J. 50 (1986) 1406. E. Sauzat, D. Guillaume, F. Villiras, J. Dubessy, M. Franois, C. Pfeiffert, M. Pelletier, R. Ruck, O. Barrs, J. Yvon, M. Cathelineau, Caractrisation Minralogique, Cristallochimique et Texturale de lArgile MX-80, ANDRA, 2001. G. Sposito, K.M. Holtzclaw, C.T. Johnston, C.S. Le Vesque, Soil Sci. Soc. Am. J. 45 (1981) 1079. G. Sposito, K.M. Holtzclaw, L. Charlet, C. Jouany, A.L. Page, Soil Sci. Soc. Am. J. 47 (1983) 51. M. Schlegel, A. Manceau, D.L. Chateigner, L. Charlet, J. Colloid Interface Sci. 215 (1999) 140. G. Heron, C. Crouzet, A.C.M. Bourg, T.H. Christensen, Environ. Sci. Technol. 28 (1994) 1698. C. Tournassat, A. Neaman, F. Villiras, D. Bosbach, L. Charlet, Am. Mineral. 88 (2003) 1989. Y. Besnus, R. Rouault, Analusis 2 (1973) 111. J. Samuel, R. Rouault, Y. Besnus, Analusis 13 (1985) 312. R.J. Hunter, M. James, Clays Clay Miner. 40 (1992) 644. M.H. Bradbury, B. Baeyens, Geochim. Cosmochim. Acta 63 (1999) 325. M.H. Bradbury, B. Baeyens, Geochim. Cosmochim. Acta 66 (2002) 2325. D.L. Parkhurst, C.A.J. Appelo, Phreeqc2 Users Manual and Program, U.S. Geological Survey, 1999, available at: http://www.geo.vu./ nlusers/posv/phreeqc.html. F. Claret, A. Bauer, T. Schfer, L. Griffault, B. Lanson, Clays Clay Miner. 50 (2002) 632. M. Gilbert, H. Laudelout, Soil Sci. 100 (1965) 157. A. Maes, P. Peigneur, A. Cremers, in: International Clay Conference, 1975. A.J. Metson, Neo-Zealand Soil Buro Bull. 12 (1956).