Vermiculite Modified with Fe 3+ Polyhydroxy Cations Is a Low-Cost and Highly Available Adsorbent for the Removal of Phosphate Ions

: This paper demonstrates that intercalating Na + homoionic vermiculite with Fe 3+ polyhydroxy cations (1:1 molar ratio OH − to Fe 3+ ) significantly improved the affinity of the clay mineral-based sorbent toward phosphate. Kinetic experiments revealed that adsorption is fast, approaching an equilibrium within about 200 min of contact time, and that the rate-limiting step is the intraparticle diffusion. Adsorption isotherms fitted to the Freundlich equation and a two-site Langmuir model, consistent with the heterogeneity of adsorption sites. The separation factor derived from the Langmuir constant revealed that the adsorption was favorable and even irreversible for high-affinity minor adsorption sites. The adsorption capacity was 299 ± 63 μmol·g −1 (9.3 ± 2.1 mg P·g −1 ), a value similar to several other clay-based phosphate adsorbents. Application to reservoir water spiked with 10 mg·L −1 in P removed about 71% of the available phosphate.


Introduction
Phosphate is a limiting nutrient for microalgae growth in aquatic environments. A lack of phosphate in wastewaters, municipal wastes, and sewage sludges implies microalgae blooms that result in the severe eutrophication of water bodies [1]. For instance, phosphate concentrations as low as 1.12 μmol·L −1 (0.11 mg·L −1 PO4 3− ) may be related to phytoplankton blooms in Brazilian urban streams [1]. There is a quest to reduce phosphate emissions or develop materials that efficiently remove phosphate from wastewaters to prevent it from reaching lakes, rivers, and reservoirs [2][3][4][5][6].
For instance, modifying bentonites (a mixture of clay minerals containing 60% to 80% montmorillonite, Mt) with La 3+ has been the basis for producing commercially available sorbents for phosphate (Phoslock ® ). The phosphate uptake occurs by forming LaPO4, which precipitates on the surface of the clay support [10,12,22,23].
The retention of oxyanions onto Fe 3+ polyhydroxy cations Mt and Vt may be explained by mixed mechanisms involving electrostatic interactions with the remaining positive charges on the partially hydrolyzed cations, or through a ligand exchange, as demonstrated by the pH variations resulting from the adsorption [21,24,25]. Furthermore, the intercalation of polyhydroxy cations enhances the basal spacing, specific surface area, pore volume, and pore diameter of the clay minerals.
The main differences between Mt and Vt are their external specific surface areas (SSAs) which are determined by different methods using gas (N2) or liquid (ethylene glycol monoethyl ether, EGME). While Mt affords an EGME SSA of up to 475 m 2 g −1 [26], Vt has 112 m 2 g −1 [21]. On the other hand, Vt has a CEC range of 80 to 126 cmolc kg −1 [24], whereas the CEC for Mt is between 60 and 150 cmolc kg −1 [22]. Another difference is that the basal spacing (d001) of Mt is systematically larger than that of Vt because its layer charge allows the separation between layers to occur at larger dimensions in an aqueous medium, producing materials with swelling properties [22].
Our research group studied Vt and Mt as adsorbents for aquatic pollutants removal [23,25,[27][28][29][30][31]. In recent work, we studied phosphate removal by Na + homionic Mt prepared using its K10 commercial form before and after intercalation with Fe 3+ polyhydroxy cations [19]. The present work demonstrates that intercalating Na + homoionic Vt with Fe 3+ polyhydroxy cations is a simple process that results in an efficient adsorbent for phosphate anions with an adsorption capacity similar to several other materials, including Phoslock ® , but using low-cost Fe 3+ salts for the chemical modification.

Preparation and Characterization of the Adsorbents
Before modification with the Fe 3+ polyhydroxy cations, the Na + -exchanged Vt (Na-Vt) was obtained as previously described [31]. The Fe 3+ polyhydroxy cations suspension was obtained by adding 50 mL of 0.40 mol·L −1 NaOH to 50 mL of 0.40 mol·L −1 FeCl3 at the flow rate of 1.0 mL min −1 under vigorous stirring, providing a 1:1 molar ratio of OH − to Fe 3+ . The intercalation suspension was maintained at 50 °C for 48 h. After this incubation time, the intercalating suspension was pumped to the Na-Vt dispersion (heated at 50 °C) under vigorous stirring at 1.0 mL min -1 using a peristaltic pump. This procedure produced a ratio of 10 mmol of Fe 3+ per gram of Na-Vt. Next, the suspensions were left standing for 72 h. The suspensions were centrifuged at 1000× g for 10 min in a Sorvall ST16R centrifuge (Thermo Fisher Scientific, Waltham, MA, USA), and the solid was washed five times with deionized water. Modified Vt was dried at 40 °C for 12 h under vacuum, crushed, and stored in a desiccator. It was named FeOH-Vt.
The basal spacing of Vt and FeOH-Vt was estimated by X-ray diffraction (XRD) in a Rigaku Miniflex instrument using Cu Kα radiation source at voltage = 30 kV, current = 15 mA, scattering slit = 4.2° and receiving slit = 0.3 mm. Continuous scan mode was used at scan speed = 1.000 min −1 , sampling width = 0.020°, and scan range from 2.0 to 50.000. Scanning Electron Microscopy (SEM) was conducted with a Fesem Jeol JSM -740 1 F instrument (Jeol Ltda, Tokyo, Japan).
The cation exchange capacity (CEC) of Na-Vt was determined by the exchange of sodium-saturated vermiculite with 1.0 mol·L −1 ammonium acetate (pH 7.0) as described in [34]. The released Na + was quantified by flame photometry.

Adsorption Studies
Adsorption experiments used a thermostatic (25.0 ± 0.5 °C) orbital shaker at the stirring rate of 250 r.p.m. All the adsorption experiments were conducted in polypropylene centrifuge tubes from Corning ® (capacity of 15 mL), horizontally accommodated inside the shaker. Dispersions were centrifuged at 4000× g for 15 min and filtered through 0.45-μm syringe filters (hydrophilic Minisart ® , Sartorius Stedim Biotech GmbH, Germany).
The effect of the adsorbent dose evaluated the adsorption of 10.0 mL of 4.84 × 10 2 μmol·L −1 (15 mg·L −1 in P) phosphate buffer (pH 6.9) onto 25.0, 50.0, 75.0, and 100 mg of adsorbent dispersed in 10.0 mL of deionized water. The contact times were 200 min (data available in Table S1).
Kinetic curves were obtained by dispersing 2.0 g of FeOH-Vt in 200 mL of 4.84 × 10 2 μmol·L −1 phosphate solution in a 500 mL beaker. The dispersion was continuously stirred in the orbital shaker, and small dispersion aliquots (1 mL) were sampled after 5, 10, 20, 40, 80, 120, 160, 200, 240, 280, 320, and 360 min of contact time (data available in Table S2). The dispersions were then quickly filtered through 0.45-μm syringe filters, and the solutions were stored at 4 °C until the determination of the free phosphate concentrations.
Adsorption isotherms (25.0 ± 0.5 °C) were constructed by adopting 200 min of contact time, using 100 mg of adsorbent dispersed in phosphate solutions (10.0 mL, pH 6.9), with the initial concentrations varying from 32.3 to 8.07 × 10 3 μmol·L −1 (1.0-250 mg·L −1 in P). The dispersions were centrifuged at 4000× g, and the supernatants were carefully separated, avoiding resuspension of the centrifuged adsorbent. The supernatants were filtered through syringe filters and stored in the refrigerator (4 °C) until the analysis (data available in Table S3).
The spectrophotometric determination of the free phosphate concentrations was based on their reaction with ammonium molybdenum (2.5 × 10 −3 mol·L −1 in 0.20 mol·L −1 HNO3) followed by the reduction of Mo(VI) to Mo(V) by 5.0% (m v −1 ) ascorbic acid in the presence of 100 mg·L −1 Sb as a catalyst [35]. Absorbance measurements were made in an Ocean Optics USB4000 spectrometer (Ocean Insight, Rochester, New York, USA) with an LS-1-LL tungsten halogen light source, connected by two optical fibers (600 μm diameter, 1 m long) to a 1 cm pathlength flow cell.

Computation
The adsorbed amount of phosphate (q) was computed as: where c , is the initial concentration of phosphate (in μmol·L −1 ), c − is the free concentration after a given contact time with the adsorbent, V is the volume of solution where the adsorbents were dispersed (10.0 mL), and m is the mass of the adsorbent.

Adsorption Kinetics
The experimental data were fitted by pseudo first order (PFO), pseudo second order (PSO), and intraparticle diffusion (ID) kinetic models, given by Equations (2), (3) and (4), respectively [36][37][38][39]: where qt is the amount of phosphate (μmol) adsorbed per gram of adsorbent at time t, qe is the fitted amount of phosphate adsorbed at the equilibrium (μmol·g −1 ), and k1, k2, and kID are the pseudo first order, pseudo second order, and intraparticle diffusion rate constants, respectively. The term C (Equation (4)) is related to the thickness of the boundary layer.
where qmax is the maximum amount of phosphate adsorbed per unit mass of FeOH-Vt (μmol·g −1 ), and KL is the Langmuir adsorption constant (L μmol −1 ). In Equation (6), the indexes 1 and 2 correspond to sites 1 and 2, respectively. From the KL, it is possible to calculate the dimensionless constant separation factor (RL) using Equation (7): where C0 is the largest initial concentration of adsorbate used in the adsorption isotherm (in μmol·L −1 ). If 0 < RL < 1, the adsorption is considered favorable, if RL > 1 the adsorption is unfavorable and if RL = 0 the adsorption is irreversible [40]. The Freundlich equation is represented by: where Kf is the Freundlich empirical constant related to adsorption capacity (μmol 1−1/n g −1 L 1/n ), and 1/n is the nonlinearity parameter associated with the energetic heterogeneity of the adsorption sites.

Data Fitting
All fittings were made using the Origin 2020 64-bit Academic software (OriginLab Corporation, Northampton, MA, USA). While the one-site Langmuir and Freundlich equations were in the library of equations as Power Origin Functions, the code of the twosite model was quickly added to the equations library. Fittings were made using the Levenberger-Marquardt iteration algorithm. The maximum number of iterations and tolerance were 400 and 1 × 10 −9 as the software's default. The fitting quality was tested by calculating the coefficient of determination (R 2 , Equation (9)) and the chi-squared parameter (χ ) (Equation (10)).
where qe,exp is the experimental value of q, measured at equilibrium, qe,calc is the fitted value of q, and qe,mean is the mean value of experimental q. The closer R 2 is to the 1, the better the fitting quality. In Equation (10), if qe,calc using a model is similar to the qe,exp, χ is close to zero. High χ values indicate high bias between the experimental data and the tested model.

Characterization of Na-Vt and FeOH-Vt
The CEC of Na-Vt was 1.154 ± 0.006 mmol·g −1 (115.4 cmolc·kg −1 ), a value consistent with the CEC of other vermiculites (80 and 126 cmolc·kg −1 ) determined by the Cd 2+ exchange method [24]. The iron content of the FeOH-VT was 1.208 mmol·g −1 , as described by Nascimento and Masini [21]. This value approaches the CEC, suggesting the Fe 3+ polyhydroxy cations quantitatively exchanged the Na + in the modification process. The incorporation of Fe 3+ polycations changed the adsorbent color from gray to golden brown (Figure 1). The X-ray diffractometry of Na-Vt exhibits the typical peak at 2theta = 6.24° relative to the basal spacing (d001) of 1.415 nm (Figure 2), characteristic of hydrated vermiculites exchanged with Na + . Other typical 2theta reflections appear at 24.16° and 30.28°, in agreement with Marcos et al. [41]. Other reflections at 2theta = 7.40°, 7.94° may be assigned to interlayered biotite [24], whereas reflections at 28.28° and 29.28° are due to unknown impurities. The exchange of Na + by Fe 3+ polyhydroxy cations did not promote any significant alteration in the d001 basal space of the modified vermiculite ( Figure 2). Similarly, the total surface area determined by the EGME method did not suffer significant alterations (122 ± 12 m 2 ·g −1 for FeOH-Vt and 112 ± 3 m 2 ·g −1 for Na-Vt) [21,31].

Dose Effect
Initial exploratory experiments demonstrated that the adsorption onto NaVt was negligible, so the following parts of the article discuss only the adsorption on FeOH-Vt. The insignificant adsorption on NaVt may be explained by the high solubility of Na + phosphate salts, contrary to Fe 3+ , which is known to form insoluble FePO4. Furthermore, the Fe 3+ polyhydroxy cations prepared with equimolar amounts of Fe 3+ and OHhave an excess of two positive charges per mole of polycations. Thus, electrostatic interaction is a possible retention mechanism in FeOH-Vt that is not likely to occur in NaVt.
The effect of the adsorbent dose for a contact time of 200 min ( Figure 4) showed that the amount of phosphate adsorbed per unit of mass increased up to 75 mg of FeOH-Vt, but decreased as the adsorbent mass increased to 100 mg. The same initial amount of phosphate (3.77 μmol, or 0.15 mg of P) distributed in a larger adsorbent mass explains the drop of q as the mass increased from 75 to 100 mg. On the other hand, the absolute amount of phosphate, or the adsorption percentage, continuously increases with the FeOH-Vt mass ( Figure 4). Thus, the adsorption kinetics and isotherms were investigated using 100 mg of FeOH-Vt dispersed in 10 mL of phosphate solution.

Adsorption Kinetics
The contact time of up 300-350 min removes about 90% of the phosphate from 10.0 mL of a 4.84 × 10 2 μmol·L −1 phosphate solution (15.0 mg·L −1 in P), dispersing 100 mg of FeOH-Vt ( Figure 5) when the system approaches the equilibrium. Fitting the experimental data with the PFO, PSO and ID models ( Figure 6) reveals that the PSO model provides slightly better R 2 and χ 2 values than those found from PFO ( Table 1). The qe was consistent in both models, at around 40-50 μmol·g −1 (Table 1). However, the kinetics limiting factor is the intraparticle diffusion (inset in Figure 6) since the plot of qt versus t 0.5 goes through the origin and has an excellent linear correlation coefficient (R 2 = 0.998, Table 1), consistent with the incorporation of Fe 3+ polyhydroxy cations in the interlamellar space of vermiculite.

Adsorption Isotherm
Adsorption isotherm data were fitted by Langmuir (one-and two-site) and Freundlich models (Figure 7). The Freundlich and two-site Langmuir models fitted the data much better than the one-site Langmuir model, as can be noticed by the R 2 , which is much closer to the unity, and the smaller values of χ 2 ( Table 2). The maximum adsorption capacity (qmax,1 + qmax,2 = 299 ± 69 μmol·g −1 , or 9.3 ± 2.1 mg P·g −1 ) is comparable with several other phosphate adsorbents described in the literature, as will be discussed in the next section of this paper.
The dimensionless separation factor RL, computed for a C0 = 8.07 × 10 3 μmol·L −1 , was between 0.15 and 0.29, assuming the one-site Langmuir model, thus implying favorable adsorption. For the two-site model, the RL values were between 0.0041 and 0.012, indicating irreversible adsorption on site 1, whereas for site 2, the RL was within the 0.45-0.71 range, thus indicating favorable adsorption.

Discussion
FeOH-Vt adsorbed phosphate from an aqueous medium with a performance similar to other adsorbents already described. For instance, the equilibrium approached with a contact time of about 200 min, corresponded with a similar condition reported for Phoslock™ [5], which required three hours of contact time. Other adsorbents such as Al/Fe-substituted bentonite required much more time (about nine hours) [42]. Other bentonite-based adsorbents prepared with La 3+ /Al 3+ or La 3+ /Fe 3+ substitutions required from 12 [43] to 96 h [11] to approach equilibrium. Several modified bentonites reached equilibrium within 24 h of contact time [44][45][46][47][48]. Few studies used vermiculite, but Huang et al. investigated the adsorption of phosphate on La(OH)3-modified exfoliated vermiculite, finding an equilibrium with about 48 h of contact time [47].
Our research group recently modified commercial K10 montmorillonite with Fe 3+ polyhydroxy cations prepared with different Fe:OH − molar ratios, including the 1:1 ratio used to modify Vt [19]. We named that material Mt-OH/Fe1:1 and studied the phosphate adsorption under the same experimental conditions used for FeOH-Vt. Compared to the modified acid-activated K10 montmorillonite, the FeOH-Vt exhibited lower rate constants (0.012 ± 0.001 against 0.12 ± 0.01 min −1 in the PFO model, and (2.4 ± 0.3) × 10 −4 against 0.010 ± 0.001 g min μmol -1 for the PSO model). The faster adsorption onto Mt-OH/Fe1:1 may be explained by the acid activation process that delaminates the structure and dissolves individual platelets leading to the formation of a low crystallinity, highly porous hydrated phase [49]. Thus, the phosphate adsorption on Mt-OH/Fe1:1 is predominantly an external surface process, corroborated by applying the ID model (Equation (4)), which resulted in a C term of 24 ± 1 μmol·g −1 , close to the qe values fitted by nonlinear regression (around 28 μmol·g −1 ) [19]. In contrast with the Mt-OH/Fe1:1, the ID model applied to the adsorption on FeOH-Vt resulted in a C value indistinguishable from zero (Table 1) and an excellent linear fitting in the time window between 0 and 180 min, thus suggesting that the modification of Vt occurred in the interlayers, and the pore diffusion became the rate-limiting step controlling the adsorption.
A one-site Langmuir model poorly fitted the adsorption isotherms, but the fitting was much improved using the Freundlich equation or the two-site Langmuir model. This finding is consistent with adsorption onto heterogeneous adsorbents, as corroborated by the 1/n Freundlich parameter of 0.47 ± 0.02 and the presence of adsorption sites with diverse adsorption strengths suggested by the excellent fitting to the two-site Langmuir model. The total adsorption capacity is 299 ± 69 μmol·g -1 , which is similar to the results reported for Phoslock™ and several other adsorbents such as Fe/Zennith bentonite (360 μmol·g −1 [5]), Al/Fe bentonite (181-364 μmol·g −1 [42]), La/Al Mt (332-420 μmol·g −1 [43]), and La/Fe Mt (30.1-188 μmol·g −1 [11]), but being prepared with low-cost Fe 3+ salts, rather than La 3+ compounds, it is much more expensive, especially considering the use in large scale.
The qmax for FeOH-Vt is smaller than that found for Zr-modified bentonite (292.5-432.6 μmol·g −1 [50]) and Zr/Al bentonite (555 μmol·g −1 [48] and significantly lower than that observed for La(OH)3-modified exfoliated vermiculite (2.57 × 10 3 μmol·g −1 [47]). The increased adsorption capacity of Zr-modified bentonites may be related to the tetravalent nature of the cation and thus a more significant positive charge density on the external surface and an increase in the basal space, pore volume, and specific surface area compared to the unmodified Mt [48]. Regarding La(OH)3-modified exfoliated vermiculite, the impressive increase in qmax may be related to the increase in specific surface area, pore volume, and pore diameter [47], which were not evidenced in the FeOH-Vt material compared to unmodified Na-Vt.
Compared with Mt-OH/Fe1:1 under the same experimental conditions, the qmax of FeOH-Vt is significantly higher (299 ± 69 against 69 ± 3 μmol·g −1 ). The higher qmax strongly correlates with the iron content in the FeOH-Vt (1.208 mmol·g −1 ) compared to that of Mt-OH/Fe1:1 (0.57 ± 0.03 μmol·g −1 ) [19], providing a greater abundance of adsorption sites to interact with phosphate ions via the precipitation of isolube iron-phosphate salts, or electrostatic interactions with the exceeding positive charges of the polycations.
The separation factor (RL) determined from the one-site Langmuir KL (0.15 < RL < 0.29) indicates that the adsorption is favorable. Regarding the two-site model, a similar finding comes from the RL between 0.45 and 0.71 obtained for site 2, which is more abundant, but has a low KL, that is, a low energy process, while for site 1, the RL was in the 0.0041 to 0.012 range, indicating an irreversible process in a highly energetic interaction with sites of low abundance. Thus, whereas the adsorption is favorable in all the studied phosphate concentration ranges, the adsorption is irreversible under the lower concentrations interacting with site 1. This heterogeneity of adsorption sites is consistent with the excellent fitting obtained with the Freundlich equation, especially with the 1/n value of 0.47 ± 0.02, characteristic of highly heterogeneous adsorption processes [37,51].
FeOH-Vt was then applied to a water sample collected at the Guarapiranga Reservoir, which plays an essential role as a water supply to about four million people in the Metropolitan Area of São Paulo city, Brazil. Although subject to the diffuse discharge of municipal waste, no phosphate was quantifiable by the spectrophotometric blue molybdenum method (LOQ = 2.0 μmol·L −1 , or 0.063 mg·L −1 in P). Thus, the sample was spiked with 251 μmol·L −1 phosphate (10 mg·L −1 in P) and exposed to FeOH-Vt under constant stirring for 200 min. After this contact time, the remaining phosphate concentration was 72.3 μmol·L −1 (2.9 mg·L −1 in P), implying about 71% removal.

Conclusions
Modifying vermiculite with Fe 3+ polyhydroxy cations enhanced the capacity and affinity of the obtained FeOH-Vt compared to the unmodified clay mineral for phosphate removal. The modification is simple, using low-cost, environmentally compatible Fe 3+ salts. Phosphate adsorption was fast compared to other adsorbents and limited mainly by intraparticle diffusion. The adsorption was favorable, exhibiting irreversible interaction in a small fraction of the adsorption sites as suggested by fitting the data to the two-site Langmuir isotherm.

Data Availability Statement:
The data presented in this study are available in the supplementary material.

Conflicts of Interest:
The authors declare no conflict of interest.