The raw brick used in this study was obtained from Bangui region in the Central African Republic. Previously, X-ray diffraction and chemical analysis were performed on this material: ~61 wt% quartz; ~21 wt% metakaolinite; 3 - 4 wt% illite; ≤4 wt% iron oxides/hydroxides; and ≤2 wt% feldspar + mica + biotite [5] . Before use, several physical/chemical treatments were carried out on the raw brick [5] [6] . Briefly: 1) it was broken into grains and sieved with sizes ranging from 0.7 to 1.0 mm ; 2) the resulting particles were leached with a 6M HCl solution at 90˚C for 3 hours; and 3) finally a deposition of FeOOH onto HCl-treated brick was performed by precipitation of iron(III) with a NaOH solution up to reach pH 7. Previously, it was noticed that HCl concentration, reaction temperature and time had a significant influence on the final meso-structure of the material [6] . Thus, following the acid leaching procedure described above, the surface area (S.A.) was found to increase notably from S.A. = 31.2 m²∙g⁻¹ in the raw brick to S.A. = 70 m²∙g⁻¹ in the modified material.

All chemicals used in the study were analytical grades. Sodium hydroxide and hydrochloride acid were supplied by DISLAB ( France ). The following salts: NaNO₃, Cd(NO₃)₂∙4H₂O, Pb(NO₃)₂, and Sr(NO₃)₂ were purchased from Prolabo. Cu(NO₃)₂∙3H₂O was obtained from Scharlau, and Zn(NO₃)₂∙6H₂O from Fluka.

To evaluate the zeta potential of brick samples, electrophoretic mobility measurements were conducted at 298˚K using a Zetameter model Zetasizer Nano ZS90 (Malvern Instruments) equipped with a laser He-Ne at 633 nm (4 mW maxi.). The electrophoresis cell (folded capillary cell) contains electrodes which are made of solid palladium and can be cleaned physically (by using an ultrasonic bath of clean water, 30 Watts) and chemically. This Zetameter apparatus determines the electrophoretic mobility of the particles automatically and converts it to the zeta (ξ) potential by means of Smoluchowski’s formula. To prevent cross-contamination, the electrophoresis cell was carefully washed with Milli-Q water several times before zeta analysis. Suspensions were prepared in a nalgene tube by mixing 500 mg of fine-grained brick (≤63 µm) with 100 mL Mill-Q water. The pH was adjusted in the range from 3.0 to 9 before each measurement by drop-wise addition of a diluted solution of either NaOH (10⁻² M) or HCl (10⁻² M). The pH which was measured after zeta-potential analysis was recorded as the final pH.

As described previously [11] , salt-addition method was used in this work in order to assess whether modified brick has either a negative or positive surface charge in an aqueous medium. Briefly, 1 g of this brick was placed in each of 10 “80-mL” beakers containing 19.5 mL of Mill-Q water. The pH of the suspensions was adjusted with HCl (10⁻² M) or NaOH (10⁻² M) to span the expected pH_PZC value. A 0.5 mL volume of a 0.1 M NaNO₃ solution (as an inert electrolyte) was added to these suspensions. All these mixtures were equilibrated for 1 night, by shaking gently at a constant speed of 120 rpm using a mechanical shaker (Model: IKA Labortechnik KS 250 basic). The equilibrium pH was recorded and designated pH_0.0025M (0.0025 M is the final NaNO₃ concentration in the medium). Afterwards, a 0.5 mL volume of a 2 M NaNO₃ solution was added to the previous mixtures, and shaken for a few minutes. The pH was recorded and designated pH_0.052 (0.052 M is the final NaNO₃ concentration in the resulting suspension). For each beaker, ∆pH = pH_0.0525 − pH_0.0025M was calculated, and ∆pH values were plotted against pH_0.0025M in order to evaluate the point where ∆pH = 0 is the point of zero charge, PZC.

Micrographies of representative specimens of modified brick were recorded by using an environmental scanning electron microscope (ESEM, Quanta 200 FEI). Elemental analysis was performed using ESEM/EDS (ESEM, model: QUANTA-200-FEI, equipped with an Energy Dispersive X-Ray Spectrometer EDS X flash 3001 and monitored by QUANTA-400 software elaborated by Bruker). EDS measurements were carried out at 20 kV at low vacuum (1.00 Torr) and the maximum pulse throughput was 20 kcps. Different surface areas ranging from 0.5 to 3.5 mm² were targeted on brick grains and examined by ESEM/EDS. Micro-observations along different across-sections of brick samples were also performed by ESEM/EDS. Atomic quantifications and mathematical treatments were undertaken using QUANTA-400 software in order to determine the averaged elemental composition of the surface brick and to detect chemical/elemental variabilities.

Continuous flow adsorption experiments were conducted in a fixed-bed glass column with an inner diameter of 12.5 mm , a height of 25 cm , and a medium porosity sintered-pyrex disk at its bottom in order to prevent any loss of material. A bed depth of 8.5 cm (10.0 - 10.5 g) was investigated at a constant flow rate of 10 mL/min. Before being used in the experiments, about five bed volumes of Milli-Q water were passed through the column for three reasons: 1) to remove any unbound and thin particles/iron oxides/hydroxides; 2) to check the absence of soluble iron at the outlet; and 3) to confirm the stability of the FeOOH coating on brick pellets. The initial concentration of cadmium(II) in the influent was 1.79 × 10⁻⁴ mole per liter. This content was chosen in this work because it generally represents a maximum level of soluble iron (≈10 mg/L) found in ground waters from the Bangui region. The Cd(II) solution was pumped through the column at a desired flow rate by means of a peristaltic pump (Modern Labo France Type KD1170) in a down-flow mode. During this column experiment, pH was measured continuously at room temperature, and effluent samples exiting the bottom of the column were collected at different time intervals and analyzed for metal contents using ICP-AES (Inductively Coupled Plasma-Atomic Emission Spectroscopy; model Varian Pro Axial View). Flow to the column continued until the effluent Cd(II) concentration at time t (C_t) reached the influent metal concentration (C₀): C_t/C₀ ≈ 0.98. Performance of the packed bed is described in the present work using the concept of the breakthrough curve.

In energy spectra for uncoated brick (i.e., HCl-activated brick), it was observed high levels of silicon, aluminum and oxygen as these are the main elements of raw brick (not shown here). Other elements including Fe, Ti, K and Mg were also observed at lower amounts. Roughly, the detected Si comes both from quartz and clays, and Al mainly from metakaolinite [6] .

In energy spectra for coated brick (i.e., brick activated by HCl and impregnated with ferrihydrite), it was clearly observed iron signals with higher intensities, which confirmed the deposition of iron oxyhydroxide into the surface of brick (Figure 1). Moreover, the energy spectrum of a Fe-rich aggregate (like that targeted and shown in the SEM image of Figure 2) revealed in all cases the presence of an Al signal at a relatively high intensity (Figure 1). This suggested that FeOOH was attached to alumino-silicates (mainly metakaolinite). To give a better insight into the way iron oxyhydroxide is attached to the brick surface, a detailed micro-observation within an across-section of a modified brick sample was further performed (see Figure 2). The line scannings for Si, Al, Fe, and Na did show the non-uniformity of element distributions along the targeted zone: indeed, an increasing level of Fe was observed alongside the iron-oxyhydroxide aggregate (see Figure 2) whereas the level of Al decreased in this area, proving the attachment of iron preferentially to clay surfaces. In other words, FeOOH deposits―being attached to clays―contributed to “mask”, at least partially, alumino-silicates specimens and thereby to deplete the EDS intensity of the Al_(embedded) peak. In contrast, the line scanning for sodium increased significantly alongside iron oxyhydroxide grains (Figure 2), suggesting that Na atoms are strongly involved in this phase (noted º SO⁻Na⁺). It should however be noted that Na atoms were detected as well in areas outside Fe-rich aggregates, raising the possibility of their binding with other hydroxyl groups (like silanol and/or aluminol from brick) in addition to -OH functions from iron oxyhydroxides. These bindings were assumed to form during the chemical treatment of brick (i.e., during Fe³⁺-ions precipitation by adding NaOH up to pH 7, leading to ferrihydrite deposition).

From previous kinetics studies [7] [12] , we examined the evolution of divalent-cation concentration in water in interaction with the adsorbent “brick” and determined the quantity of cation adsorbed onto the brick surface with time (as obtained for Pb(II) adsorption and shown in Figure 3(a)). When the equilibrium state of this kinetic reaction was attained, the dicationic metal (Me²⁺) adsorbed on brick was in equilibrium with a residual Me²⁺ content in the liquid phase. The K_D ratio was defined as the distribution coefficient of the solute between the adsorbent and the solution in equilibrium:

where C_ads represents the amount of metal adsorbed on brick at equilibrium; and C_eq is the concentration of metal in water at equilibrium [13] -[17] . The enthalpy, ΔH⁰, and entropy, ΔS⁰, of the adsorption could be estimated

from Van’t Hoff equation:

For each studied metal, LnK_D or Ln(C_ads/C_eq) was drawn against 1/T, and the plot was found to be a straight line with a regression coefficient R² > 0.90 (as observed for Pb(II) adsorption and shown in Figure 3(b)). From the slopes and intercepts of the linearized curves, enthalpies (ΔH⁰) and entropies (ΔS⁰) of the different reactions between divalent metals (Cd²⁺, Cu²⁺, Pb²⁺, Sr²⁺, and Zn²⁺) and brick in water, were evaluated and reported in Table 1. As for Gibbs free energy, ΔG⁰, it was calculated at 298˚K from the following equation:

It is worth noting that the change in free energy for physisorption is between −20 and 0 kJ/mole, while chemisorption is in the range −80 to −400 kJ/mole [13] [18] . In our case, the Gibbs free energy, ΔG⁰, for the studied process was found to be in the range of (−1.4 to −5.31) kJ/mole (see Table 1), and consequently, the mechanism was predominantly physical adsorption. Energetic data also confirmed the feasibility of the process and the thermodynamically favorable nature of the reaction. Furthermore, the negative values of ΔG⁰ were much higher

than those found for system implicating charge transfer or sharing from adsorbent to adsorbate to form coordinate bonds with ΔG⁰ ≤ −40 kJ/mole. The ΔG⁰ values obtained in this work then agreed better with a physical- adsorption mechanism in which electrostatic interactions between dicationic metals and active brick sites were able to be implicated. It was also noticed that this change of free energy decreased with increase of temperature, indicating clearly that, with increasing temperature the adsorbed amount of divalent metals increased at equilibrium.

Moreover, the positive values of enthalpy change revealed that the adsorption process was endothermic. The magnitude of ΔH⁰ may also give an idea about the type of sorption involved in our system. Indeed, chemisorption bond strengths are generally in the range of (+84) - (+420) kJ/mole, while physisorption bond strengths are <+84 kJ/mole [19] [20] . From Table 1, the weak ΔH⁰ values are in the range of (+13.9) - (+52.9) kJ/mole, and therefore confirm that Me²⁺ adsorption by this modified brick occurs more according to a physisorption process rather than a pure chemical adsorption mechanism. As for the highly positive values of ΔS⁰ (see Table 1), they were consistent with an increase in the degrees of freedom of the implicated species at the solid-liquid interface as a result of structural modifications. These latter were assumed to occur especially when water molecules were released from hydrated metal ions and/or adsorbent surface during the migration of the adsorbate from water to brick. Such structural changes led with time to the generation of solid solutions: (Me²⁺)_x(Fe³⁺)_yOOH, as suggested recently for iron(II) bound to modified brick [8] .

To summarize, both the values of ΔG⁰ and ΔH⁰ suggested that the adsorption of metallic cations onto brick was a physisorption process.

The pH of the brick-water system determines the surface charge of the solid particles and thereby should have a relevant control over the uptake of metals. Figure 4(a) illustrates the effect of pH on the zeta potential of ferrihydrite-coated (pre-activated) brick. It was observed that the positive zeta potential decreased with increasing pH and its sign reversed at pH 3.2, which was identified as the isoelectric point (IEP). This indicated that at pH > pH_IEP the brick surface possessed negative charges in water and particularly around neutral pH. It is worth noting that the pH_IEP value obtained for modified brick is higher than that for HCl-activated brick (pH_IEP = 2.2) and lower than that for raw brick (pH_IEP = 4.0 - 4.3), however, it is found to be much lower than that for pure ferrihydrite (IEP = 7.8 [21] [22] ), see Figure 4(a).

On the other hand, the pH at which the net charge on the surface (noted PZC) was zero, was determined in this work by applying the salt-addition method, as described previously [11] . As shown in Figure 4(b), ∆pH = pH_0.0525 − pH_0.0025 was plotted against pH_0.0025M (for more details, see the experimental section 2.3), and the resulting curve permitted to attain the pH_PZC of this material: pH ~ 3.0 - 3.3. It can be noticed that this pH_PZC value is close to pH_IEP. Cation adsorption takes place favorably when the pH_PZC of oxides is below the solution pH. In our experimental applications performed in the field, the pH of treated waters and/or wastewaters were generally found to be higher than 4 and hence made modified-brick surfaces negatively charged and active for cations removal. This implied that electrostatic attractions between metallic cations and brick surfaces should intervene. In this hypothesis, in what follows we had tried to appraise the importance of electrostatic forces in this adsorption process.

In a previous work [12] , we found that kinetic data relative to the adsorption of divalent cations (Cd²⁺, Cu²⁺, Pb²⁺, Sr²⁺ and Zn²⁺) onto this modified brick fitted better the pseudo-second-order model proposed by Ho and McKay [23] [24] . Figure 5(a) shows how these rate constants, k₂, are dependent upon the acidity constant of these metals, pK_A (according to the reaction: Me²⁺ + H₂O ↔ Me(OH)⁺ + H⁺ [25] -[27] ), or vary with the starting pH of the reaction medium measured in a “water + brick” suspension at 298˚K (Figure 5(b)). As a whole, an increase in the starting pH or pK_A results in increasing metal adsorption rate, thus revealing the process to be intimately dependent upon H⁺ and OH⁻ ions. Indeed, hydrogen and hydroxyl ions have long been considered to be potential determining in the adsorption of hydrolysable metal ions at the oxide-water interface [10] . Since the charge of the brick surface seems to be affected by pH, the removal by electrostatic attraction is expected to occur in our experiments between absorbent and cation having opposite charges. In this context, the interfacial chemical reactions (protonation/deprotonation of hydroxyl groups, ≡SOH, in ferrihydrite, and partially in silanol and/or aluminol from brick metakaolinite) should contribute to develop an electrostatic potential that has to be assessed. When brick is in an equilibrium state with water, the following acid-base equilibria are assumed to intervene at the water-brick interface:

Formation and dissociation of H₃O⁺ ions from the surface hydroxyls could therefore account for the surface charge, σ, on the brick which is given by:

And the site density can be written as:

We used the electrostatic diffuse layer model (DLM) in order to express the surface charge (σ) as a function of the potential at the brick surface (ψ₀, also being known as the surface potential or total double layer potential). In this approach, the Gouy-Chapman relationship was employed as:

where ε₀ is the permittivity of a vacuum (ε₀ = 18.854 × 10^?12 C²∙J^?1∙m^?1); ε_r is the relative permittivity of water (ε_r = 78.5 at 298˚K); “I” is the ionic strength of the medium; F Faraday constant; R gas constant; T absolute temperature. As a first approximation, the ψ₀ potential of an oxide surface varies with pH as [10] :

where is the activity of potential-determining ions H⁺ (for brick oxides) in the bulk of the solution in reference to that at the state of PZC, a_H+PZC; ∆z corresponds to the change in the charge of surface groups (i.e., ∆z = +1 for protonation, and ∆z = −1 for deprotonation); R, T and F have their usual significance. In this context, we plotted the pseudo-second order kinetic rate, k₂, relative to metal-adsorption kinetics against the surface potential (ψ₀) which was measured at the beginning of the kinetic process (Figure 6). It can be seen in this figure that the higher the ψ₀ potential, the speedier the cationic species can adsorb on the negatively charged mineral surface. This finding demonstrates clearly that electrostatic attractions are implicated in this process and the rate constant is strongly related to the extent of the surface potential, ψ₀.

Consequently, attractive Coulombic force should presumably be a relevant contribution in the overall free energy change (∆G_ads) for this adsorption process. To support this hypothesis and to assess the importance of this force on the kinetic reaction, some theoretical energetic aspects of the studied system were addressed in Section 3.5.

The surface complexation model proposed by James and Healy [10] was used in the past in order to evaluate the adsorption characteristics of ions on oxide surfaces [28] -[30] . In the present work, we applied this model to determine the change of the total free energy of divalent-cation adsorption onto modified brick in water from the following equation:

where DG_Coul, DG_solv, and DG_chem represent the changes in free energy due to Coulombic (or electrostatic) effects, secondary solvation, and chemical interactions, respectively. Using the electrical double layer theory, the change of Coulombic energy can be determined from the equation [10] :

with

, where Ψ₀ is the surface potential as defined above;

, where x is the hydrated ionic radius of the cation, and R_cation and R_water are the radii of cation and water, respectively;

z_i = charge on the adsorbing ion

z = 1 for 1:1 (background electrolyte).

κ represents the Debye-Huckel reciprocal double-layer length which depends up the ionic strength, I, of the medium as:

where z_i is the charge of the metal cation and C_i is the ions concentration in water. κ was calculated for our system. We found: κ @ 1.94 × 10⁹ m⁻¹ with I @ 3.44 × 10⁻¹ mol/m³.

As for the change of free energy due to effects associated with secondary solvation, its contribution is given by [10] :

where and

The radius of water was assumed to be that of oxygen (1.38 × 10⁻¹⁰ m) [10] ; ε_bulk and ε_solid are the dielectric constants for aqueous solution (ε_bulk ≈ 78.5) and FeOOH adsorbed onto brick (ε_solid ≈ 11.70 [31] ), respectively; R is the universal gas constant (8.314 J/(mol・K)); The R_cation values used in this work were those reported by Marcus [32] . The ∆G_Coul and ∆G_solv values were reported in Table 2. As a whole, the experimental DG⁰ value was found to be close enough to the sum: ∆G_Coul + ∆G_solv, indicating that the contribution of the specific chemical interaction, DG_chem, was weak during the time interval of kinetic experiments (<1 hour). It can be concluded that both the attractive Coulombic force and the repulsive secondary solvation force contribute mostly to the overall free energy change observed during metal-adsorption kinetics.

Under dynamic conditions by using a fixed-bed column (as stated in the experimental section), the effectiveness of this modified brick in the removal of a divalent metal like Cd⁺⁺ was examined. The Cd⁺⁺ content in the column influent was fixed at 1.79 × 10⁻⁴ mol/L which corresponds to 20.1 mg of cadmium per liter. Experimental data are presented graphically on Figure 7(a). To access the adsorption performance of this material, the breakthrough time (t_b) and the exhaustive time (t_e) were determined under given operational conditions: they were

defined as the elapsed time values when the effluent Cd²⁺ concentration (C_t) attained 5% (t_b) and 95% (t_e) of the influent Cd²⁺ concentration (C₀). It can be seen in Figure 7(a) that the effluent volume at breakthrough time is ~500 mL and at exhaustive time ~1500 mL. Also, another column experiment was carried out by passing through the brick column a Cd(II) solution at the same concentration as that before, however, by adding further an inert electrolyte: NaNO₃ (10⁻¹ M) in the aim to study the influence of an increase of the ionic strength on the adsorption (Figure 7(a)). The resulting breakthrough curve (Figure 7(a)) shows clearly an inhibition of the adsorption process as a result of the addition of NaNO₃ salt: indeed both the breakthrough time and exhaustive time decreased significantly. Such a reaction inhibition could be attributed to a charge reversal at the brick surface in highly concentrated electrolyte. It should be noted that this phenomenon had previously been observed on other oxides like: aluminum hydroxide/oxide [33] [34] , anatase [35] , and hematite particles [36] . In our case, the electrokinetic effect at high ionic strength could be due to the occupation of most surface sites of the brick by a specifically adsorbed ion that would be the Na⁺ species in excess in the influent. It was also observed that the addition of an inert electrolyte led to an increase of the medium acidity or a decrease of pH (see Figure 7(b)). Consequently, the surface potential, Ψ₀, diminishes as well with a decrease of pH according to Equation (9).

Therefore, any electrostatic forces at the water-brick interface tend to be canceled out and cation migration as well, in agreement with results obtained experimentally with fixed-bed column.