Adsorption of diclofenac on a magnetic adsorbent based on maghemite: experimental and theoretical studies

Abstract

In the present study, a magnetic adsorbent for diclofenac formed by maghemite (γ-Fe₂O₃) nanoparticles with high saturation magnetization (19.8 emu g⁻¹) and specific area (79 m² g⁻¹) was synthesized by a one-pot method through the precipitation of Fe²⁺ ions with NaOH solution followed by rapid oxidation with hydrogen peroxide. The X-ray diffraction and Mössbauer spectroscopy data confirmed that the adsorbent is formed solely by maghemite. The adsorption equilibrium time for diclofenac (C₀ = 500 mg L⁻¹) was reached after 120 min, and the kinetic data were best fitted to the pseudo-first-order model. The adsorption isotherms acquired at five different temperatures showed an increase in the maximum adsorption capacity (261 mg g⁻¹) until 298 K, but at higher temperatures, the maximum adsorption capacity was not increased. The isotherm data were best fitted to the Langmuir and Sips models. Adsorption tests as a function of solution pH showed a decrease in the diclofenac adsorption capacity with increasing solution pH, suggesting that the hydroxyl anions compete with diclofenac molecules for the adsorption sites. Diclofenac adsorption on maghemite was endothermic (67.31 kJ mol⁻¹) and entropically driven (TΔ_adsS° = 96.33 kJ mol⁻¹). Finally, theoretical calculations and infrared spectroscopy data suggest a physisorption mechanism of diclofenac on the maghemite surface.

---

Introduction

Among the various toxic organic compounds known, contaminants of emerging concern, such as pharmaceuticals and personal care products (PPCPs), have been widely studied by different research groups in recent years.^1,2 The increasing number of PPCPs in the environment has drawn the attention of governmental and non-governmental organizations worldwide due to their impact. Indeed, some international agencies have listed some pharmaceuticals and endocrine-disrupting compounds (EDCs) as priority substances or placed them on the list of potential candidates to be considered as priority substances. In the European Union, some drugs such as the contrast medium amidotrizoate, the antiepileptic carbamazepine, the fungicide clotrimazole and the anti-inflammatory diclofenac (DCF) were included in the list of compounds to be considered as priority substances, even though the approved Directive 2008/105/EC³ did not consider them as such. Recently, the Directive 2013/39/EU of the European Parliament and the Council has decided that DCF shall be included in the first watch list, to gather monitoring data to facilitate the determination of appropriate measures to address the risk posed by this substance. If approved, the provisional guideline for DCF in surface waters would be 100 ng L⁻¹.

Nevertheless, due to their chemical resistance and xenobiotic nature, it is necessary to develop useful technology for removing those compounds from contaminated wastewater^1,4 so that their impacts on aquatic life are minimized, thereby abiding to increasingly stricter legislation. For instance, the Resolution 430/2011 of the Brazilian Environmental Council (CONAMA) states that the discharge of effluents should not cause toxic effects to aquatic organisms in the receiving body, according to the ecotoxicity criteria established by the environmental agency.

A pharmaceutical that has been found in raw and treated sewage, as well as surface waters, is the nonsteroidal anti-inflammatory DCF, which is widely used for relief of pain and inflammation, especially in specific cases such as rheumatoid arthritis, osteoarthritis, tendonitis and postoperative pain.⁵ Another study showed the occurrence of three pharmaceuticals (carbamazepine, clofibric acid, and diclofenac) in treated wastewaters in four European countries and evaluated their environmental impacts toward bacteria, algae, invertebrates, and fish.⁶ The results demonstrated that all tested pharmaceuticals were detected and quantified in most of the effluents analyzed.^6,7 Diclofenac induces oxidative stress on Hyalella azteca, and the set of tests used in this study (lipid peroxidation, protein carbonyl content and antioxidant enzyme activities) is considered as an adequate early damage biomarker for use in evaluating the toxicity of that pharmaceutical group in aquatic species.⁷

Various technologies have been tested for the removal of pharmaceuticals from contaminated water. Among these, the adsorption process deserves to be highlighted since it can be more easily coupled to existing water treatment plants that employ the conventional water treatment process (coagulation, flocculation, clarification, and disinfection). In the adsorption process, the pollutant is removed from a liquid phase to the surface of an adsorbent material, and among the studied materials used to adsorb diclofenac, one should highlight porous adsorbents such as activated carbon, alumina, zeolite, silica, and iron-based materials.^8–11

Carbon-based materials are the most widely used adsorbents for DCF in water due to their high hydrophobicity and high surface area.^12–14 However, the carbon-based material powders tend to mix with the suspended solids in real wastewater, making it difficult to separate them and impossible to reuse particulate carbon-based adsorbents. On the other hand, maghemite is attractive to be used as an adsorbent because of its high saturation magnetization,¹⁵ which facilitates its separation from the adsorption medium through the application of a magnetic field for further use. Also, the abundance of iron in the environment contributes to its low-cost preparation compared to carbon materials such as carbon nanotubes and graphene.^16–18

Maghemite is usually prepared in two steps: in the first step, magnetite is synthesized by the chemical co-precipitation of ferrous and ferric ions in an alkali solution, and in the second step, magnetite is oxidized at 200–400 °C to form maghemite.¹⁵ Alternatively, different methods have been reported for the synthesis of maghemite nanoparticles such as hydrothermal,¹⁹ solvothermal,²⁰ microemulsion,²¹ and thermal conversion.²² However, these methods of synthesis have some inconvenience for large-scale preparation such as pH control, the use of surfactants, long-time reactions, and high-temperature reactions.

Herein, we present a simple, one-step, fast, room temperature, and clean method for the synthesis of maghemite nanoparticles and their use as an adsorbent for diclofenac, an emerging pollutant of water bodies. A detailed characterization of maghemite was carried out by different techniques. Kinetic and equilibrium adsorption studies were performed to understand the mechanism of DCF adsorption. Also, thermodynamic parameters and theoretical calculations were used to investigate the adsorption mechanism and the types of interactions occurring between DCF and maghemite.

Material and methods

Chemicals

Sodium diclofenac (2-[(2,6-dichlorophenyl)amino]-benzeneacetic acid sodium salt) (318.13 g mol⁻¹) was purchased from Sigma-Aldrich (Brazil). Fe(SO₄)₂(NH₄)₂·6H₂O, HCl, H₂O₂ (35%, v/v), NaCl, and NaOH were purchased from Synth (Brazil).

Adsorbent preparation

For the synthesis of γ-Fe₂O₃ nanoparticles, 5.56 g of Fe(SO₄)₂(NH₄)₂·6H₂O was weighed in a 250 mL glass Beaker and dissolved in 30 mL of distilled water. Then, 15 mL of an aqueous 2.0 mol L⁻¹ NaOH solution was slowly (100 min) added to the Fe²⁺ aqueous solution. After the formation of a green rust precipitate, 5 mL of 35% (v/v) H₂O₂ was immediately added to the Beaker under magnetic stirring (180 rpm). Then, the suspension was stirred for 30 min and then centrifuged for 10 min at 3600 rpm. The supernatant was separated from the solid and then distilled water was added to the centrifuged solid. The centrifugation process was repeated until a neutral pH of the supernatant was obtained. The solid (maghemite) was dried at 65 °C for 18 h. The maghemite was stored in an amber flask for further use.

Characterization of the adsorbent

Powder XRD data were obtained using a diffractometer (RIGAKU, model GEIGERFLEX) under CuKα radiation (λ = 1.5406 Å, 40 kV and 20 mA). The diffraction pattern was recorded from 10 to 80° 2θ (Bragg angle) at a scan rate of 0.6° min⁻¹. Particle size was calculated using the Scherrer equation²³ (eqn (1)).where D_hkl (nm) is the particle size, K (0.89) is the Scherrer constant, λ (nm) is the X-ray wavelength, β (radians) is the full-width at half-maximum (FWHM), and θ (radians) is the Bragg angle.

Mössbauer spectra were obtained using a Mössbauer spectrometer (CMTE, model MA250) with a ⁵⁷Co/Rh radiation source at room temperature (298 K) and 80 K. Nitrogen adsorption/desorption isotherms were obtained using a Quantachrome Autosorb iQ analyzer. The specific surface area was calculated using the Brunauer, Emmett, and Teller (BET)²⁴ method and the pore size distributions were determined by the Barrett, Joyner, and Halenda (BJH)²⁵ method.

The magnetic properties of maghemite were evaluated using a vibrating sample magnetometer (VSM) (LakeShore, model 7404). VSM measurements were used to measure the hysteresis curve and the parameters of saturation magnetization (the maximum magnetization achieved with an applied field), remanence (the zero field magnetization value) and coercive field (the value of the magnetic field to be applied to the material to bring the magnetization to zero).

The infrared spectra of maghemite, DCF, and DCF-loaded maghemite were recorded on an FT-IR spectrometer Alpha (Bruker Instruments) equipped with a diffuse reflectance infrared Fourier spectroscopy (DRIFTS) module with 128 scans per sample at a resolution of 2 cm⁻¹.

Determination of the point of zero charge (PZC)

The pH of the point of zero charge (pH_PZC) of the γ-Fe₂O₃ adsorbent was determined by the solid addition method.²⁶ Samples of 20.0 mg of γ-Fe₂O₃ were transferred to Erlenmeyer flasks containing 20 mL of an aqueous NaCl (0.1 mol L⁻¹) solution. The initial pH (pH_i) of the aqueous solution was previously adjusted to different values from 2 to 12 by adding drops of aqueous 0.01 mol L⁻¹ HCl or 0.01 mol L⁻¹ NaOH solutions. The flasks were shaken at 180 rpm in an orbital incubator (Solab Scientific, model SL223) and 25 °C for 24 h to reach equilibrium. After 24 h, the final pH (pH_f) of each suspension was measured using a pH meter (Metrohm) and the data obtained were plotted in a graphic representation of ΔpH (pH_i − pH_f) against pH_i.

Determination of maghemite acidity

A sample of 0.0500 g of γ-Fe₂O₃ was accurately weighed in a 100 mL Erlenmeyer flask, and 20.0 mL of an aqueous NaOH (0.01 mol L⁻¹) solution was added. After 24 h of agitation at 180 rpm and 25 °C, the solution was separated from the maghemite powder by single filtration and titrated with an aqueous HCl (0.05 mol L⁻¹) solution up to the phenolphthalein endpoint.

Adsorption studies

For adsorption studies as a function of contact time (kinetics) and initial diclofenac concentration, the initial pH of the aqueous diclofenac solutions was adjusted with drops of aqueous 0.1 mol L⁻¹ NaOH or 0.1 mol L⁻¹ HCl solutions.

Kinetic studies

Samples of 100.0 mL of a stock diclofenac (100 or 500 mg L⁻¹) solution at pH 7.0 were added to 250 mL Erlenmeyer flasks containing 100.0 mg of the adsorbent and stirred at 150 rpm and 25 ± 1 °C in an orbital shaker incubator (Solab Scientific, model SL223) for different contact times. The adsorption capacity at a time t was determined by measuring the equilibrium concentration (C_e) of diclofenac using a UV-Vis spectrophotometer (Hewlett-Packard, model 8453) at its maximum absorbance wavelength (275 nm). A standard calibration curve was built for the quantification of diclofenac in aqueous solution (Fig. S1, ESI†).

The adsorption capacity was calculated using eqn (2):

where q (mg g⁻¹) is the adsorption capacity, C_i (mg L⁻¹) is the initial diclofenac concentration, C_t (mg L⁻¹) is the diclofenac concentration at time t, V (L) is the volume of diclofenac solution, and m (g) is the mass of maghemite.

The kinetic models commonly used for fitting the adsorption kinetic data in the liquid phase are the pseudo-first-order,²⁷ pseudo-second-order,²⁸ and intraparticle diffusion models.²⁹ The pseudo-first-order kinetic model is defined in eqn (3):

where q_t (mg g⁻¹) is the amount of adsorbed solute at a time t (min), q_e (mg g⁻¹) is the amount of adsorbed solute at equilibrium, and k₁ (min⁻¹) is the pseudo-first-order rate constant. This model assumes that the adsorption of an adsorbate is proportional to the difference between the amount of adsorbed solute at equilibrium and the amount adsorbed at time t. The initial adsorption rate (h) for the pseudo-first-order kinetic model can be calculated using eqn (4):

h/(mg g⁻¹ min⁻¹) = k₁q_e (4)

The pseudo-second-order kinetic model implies that the adsorption capacity is proportional to the number of active sites on the adsorbent surface.²⁸ The pseudo-second-order kinetic model is defined using eqn (5):

where k₂ (g mg⁻¹ min⁻¹) is the pseudo-second-order rate constant. The initial adsorption rate (h) for the pseudo-second-order kinetic model can be calculated in eqn (6):

h/(mg g⁻¹ min⁻¹) = k₂q_e² (6)

Finally, the intraparticle diffusion (IPD) kinetic model describes a mechanism in which there may be one or more steps in the adsorption on a given adsorbate. First, the adsorbate approximates to the surface of the adsorbent and then it undergoes diffusion into the pores of the adsorbent (intraparticle diffusion). Subsequently, the adsorbate is adsorbed in the inner pores of the adsorbent.³⁰ The IPD model is represented in eqn (7):

q_t/(mg g⁻¹) = k_it^0.5 + C (7)

where k_i (mg g⁻¹ min^−1/2) is the intraparticle diffusion rate constant, and C is the intercept, which is related to the thickness of the boundary layer.

Adsorption isotherms

Adsorption isotherms were performed at 25 ± 1 °C and at an initial pH of 7.0 ± 0.3. Samples of 100.0 mL of a stock diclofenac solution were added to 250 mL Erlenmeyer flasks containing 100.0 mg of adsorbent. The flasks were stirred at 150 rpm for 24 h in an orbital shaker incubator (Solab Scientific, model SL223).

The isotherm models of the Langmuir,³¹ Freundlich,³² Temkin,³³ and Sips³⁴ were used to fit the equilibrium data of diclofenac adsorption onto γ-Fe₂O₃. The Langmuir model (eqn (8)) assumes the formation of an adsorbed monolayer in which all adsorption sites present the same energy. This model also assumes that the active sites are finite and that there are no lateral interactions between the adsorbates.

In eqn (8), q_e (mg g⁻¹) is the amount of adsorbed adsorbate at equilibrium, C_e (mg L⁻¹) is the equilibrium concentration of the adsorbate in the liquid phase, Q_max (mg g⁻¹) is the maximum adsorption capacity, and b (L mg⁻¹) is the Langmuir binding constant.

The Freundlich model assumes multilayer adsorption onto heterogeneous surfaces. It is commonly described using eqn (9):

q_e/(mg g⁻¹) = K_FC^1/n_e (9)

where K_F [(mg g⁻¹/L mg⁻¹)^1/n] is the Freundlich constant, which is related to the adsorption capacity, while 1/n characterizes the adsorption intensity and indicates whether the adsorption is favorable (1/n < 1) or unfavorable (1/n > 1).

The Temkin model (eqn (10)) assumes the interactions between the adsorbent and adsorbate without having the limitation of low or high concentrations of adsorbate. It assumes that the heat of adsorption of all adsorbates decreases linearly with the adsorbent coverage and that the adsorption is characterized by a uniform distribution of the binding energies up to a maximum binding energy.³⁵

where R (8.314 J mol⁻¹ K⁻¹) is the gas constant, T (K) is the temperature, and b_T (J mol⁻¹) is the Temkin constant.

The Sips or Langmuir–Freundlich model (eqn (11)) is a model derived from the Langmuir and Freundlich equations. The adsorption system will be more heterogeneous if the value of 1/n is far from unity. On the other hand, values of 1/n close to unity also suggest homogeneous surfaces, with the experimental data approaching the Langmuir model assumptions.

where b (L mg⁻¹) is the Sips constant related to the affinity of the adsorption sites for an adsorbate.

Thermodynamic calculations

The change in enthalpy of adsorption (Δ_adsH°) and the change in entropy of adsorption (Δ_adsS°) were calculated using the van’t Hoff equation (eqn (12)) as follows:³⁶where K_a (dimensionless) is the thermodynamic equilibrium constant. The thermodynamic equilibrium constant can be calculated from the Langmuir constant using the approaches suggested by Liu³⁷ and Ghosal and Gupta³⁸ as follows:where γ_e (dimensionless) (eqn (13)) is the activity coefficient at equilibrium, which was calculated by taking the first diclofenac equilibrium concentration of the plateau of the adsorption isotherms at different temperatures and b (L mol⁻¹) is the Langmuir constant.where z is the charge of the ion in solution, which is equal to −1 for diclofenac at pH 7.0 and μ (mol L⁻¹) is the ionic strength of the solution. The ionic strength of the solution can be calculated by the Guntelberg approximation³⁹ from the extended Debye–Hückel law and applies to ionic strengths up to 0.1 mol L⁻¹.

The change in free energy of adsorption (Δ_adsG°) was calculated using eqn (15) as follows:

Δ_adsG°/(kJ mol⁻¹) = Δ_adsH° − TΔ_adsS° (15)

Computational methods

The DFT (Density Functional Theory) method was used for all calculations with the software ADF-BAND 2009.01.^40,41 The maghemite surface and the diclofenac molecule were studied for a better understanding of how they interact.

According to Yang et al.,⁴² the accuracy of the DFT derived results depends on the choice of functional and basis sets. The exchange–correlation functional in the generalized gradient approximation (GGA) is expressed using both the local electron density and the gradient of the electron density. Thus, it is held that the GGA functionals can provide more accurate results.⁴²

The calculations for the optimization of the structures were performed using the PBE/GGA functional. This functional has been used with the TZP basis set, which is a large uncontracted set of Slater-type orbitals containing diffuse functions, which is of triple-ζ quality and has been improved with one polarization functional set: 3d on carbon, 4f on iron, and 2p on hydrogen. The parameters used to build the maghemite structure were based on previous studies,^49–51 with the space group Fd3m as the cubic structure. The structure of maghemite with the space group Fd3m has both Fe³⁺ atoms per formula unit at tetrahedral sites “0.125, 0.125, and 0.125” and octahedral sites “0.5, 0.5 and 0.5”, and the positions of O atoms are defined by their coordinates “0.25468, 0.25468 and 0.25468”, with the lattice parameter a = 0.834 Å.

Results and discussion

Characterization of the adsorbent

In this study, a new, one-step, low-cost and straightforward method was used to synthesize maghemite. Table 1 provides a short comparison of the methods reported for synthesizing maghemite. The maghemite produced in the present study has a high surface area, similar to those produced by different methods, but with the advantage of using a green, simple, and fast method at room temperature and without using surfactants.

Table 1 Synthesis methods for maghemite preparation

Synthesis method	Surface area (m^2 g^−1)	Ref.
Chemical method	31.0	43
Chemical method	73.8	44
Co-Precipitation method	81.61	45
Sol–gel method	198.0	46
Flame spray pyrolysis	79.3	47
Maghemite NPs (Sigma-Aldrich)	40.8	48
Chemical precipitation	79.0	This study

---

The powder XRD pattern of the synthesized adsorbent is shown in Fig. 1a. The diffractogram of the synthesized material shows peaks that are characteristic of cubic maghemite, according to the Joint Committee on Powder Diffraction Standards (JCPDS), File No. 24-81, thus confirming that γ-Fe₂O₃ was successfully produced. The crystallite size of maghemite calculated using the Scherrer equation (eqn (1)) was estimated to be 10 nm.

Fig. 1 (a) Powder XRD pattern of the γ-Fe₂O₃ adsorbent. (b) Mössbauer spectra of the γ-Fe₂O₃ adsorbent at 298 K and 80 K.

The ⁵⁷Fe Mössbauer spectra of the synthesized maghemite recorded at 298 K and 80 K are shown in Fig. 1b. The spectrum of the sample recorded at 298 K (Fig. 1b) shows a magnetically ordered component (sextet), with the hyperfine field at approximately 48.5 T and a quadrupole shift at around zero, which is consistent with the formation of maghemite, as confirmed by XRD. Also, an intense doublet with an isomer shift of 0.36 mm s⁻¹, typical of Fe³⁺ ions in octahedral coordination, was observed. This doublet suggests that the iron species have small particle size. To confirm this hypothesis, Mössbauer measurements were collected at 80 K, and the obtained spectrum is also shown in Fig. 1b. The spectrum was fitted with two subspectra corresponding to Fe³⁺ in octahedral (solid red lines) and tetrahedral (short dashed blue lines) sites in the maghemite structure. In addition to these two sextets, a Fe³⁺ doublet can still be noted even at 80 K, reflecting that the sample is constituted by two principal distributions of particle sizes: (i) larger particles (sextet in the Mössbauer spectrum) and (ii) smaller particles (doublet).

Fig. 2a shows the N₂ adsorption–desorption isotherms for the γ-Fe₂O₃ adsorbent, which exhibited a type IV isotherm that is characteristic of mesoporous materials and a type H1 hysteresis loop that is related to the filling of the mesopores due to capillary condensation.⁵¹ The specific surface area of the γ-Fe₂O₃ adsorbent was 79 m² g⁻¹, and the pore size distribution curve (Fig. 2b) obtained by the BJH method showed a relatively narrow pore distribution in the range from 2 to 12 nm, thus confirming the formation of mesopores in the γ-Fe₂O₃ adsorbent. The synthesized maghemite exhibited mesopore and micropore volumes of 1.199 cm³ g⁻¹ and 0.056 cm³ g⁻¹, respectively.

Fig. 2 (a) Nitrogen adsorption–desorption isotherms and (b) the pore size distribution for the γ-Fe₂O₃ adsorbent.

The maghemite acidity was found to be 0.25 mmol g⁻¹, indicating that the adsorbent shows low surface acidity.⁵² The acidity can be related to the hydroxyl groups on the maghemite surface, which may also contribute to the adsorption process.

The magnetization curve for the γ-Fe₂O₃ is shown in Fig. S2 in the ESI.† The corresponding saturation magnetization of the synthesized maghemite nanoparticles was 19.8 emu g⁻¹. This magnetization value for maghemite nanoparticles was higher than that reported by Hu et al.⁴⁶ It is worth noting that after the application of the external magnetic field, a complete magnetic separation of the maghemite was achieved in 60 s by a permanent magnet near the wide-mouth bottle (Fig. S2, inset, ESI†). This property can be used for recovering the adsorbent from the adsorption medium for further reuse.

The infrared (IR) spectrum of maghemite (Fig. S3, ESI†) shows various bands commonly observed for iron oxides, which are related to the vibrations of Fe–O and Fe–O–H bonds. The band at 567 cm⁻¹ was attributed to stretching of Fe–O while the bands at 896 cm⁻¹ and 795 cm⁻¹ were attributed to the bending of the hydrogen bond of Fe–OH⋯H. The broadband between 1442 and 1644 cm⁻¹ was related to the angular bending of adsorbed water molecules. Finally, the region between 3000 and 3500 cm⁻¹ indicated the presence of hydroxyl groups and also the stretching vibrations of adsorbed water molecules on the maghemite surface.⁴⁷

Adsorption experiments

Kinetic studies. The removal of diclofenac from aqueous solution using the γ-Fe₂O₃ adsorbent as a function of contact time is shown in Fig. 3a. The adsorption kinetics was performed using two different initial concentrations (100 and 500 mg L⁻¹) of diclofenac aqueous solution to evaluate the effect of the initial concentration on the kinetic parameters, i.e., the equilibrium adsorption capacity (q_e) and the adsorption rate constant (k), as well as the mechanism of adsorption.

Fig. 3 (a) Adsorption of diclofenac on γ-Fe₂O₃ as a function of time at different initial diclofenac concentrations at 298 K (25 °C). (b) Intraparticle diffusion plots for the kinetic data of diclofenac adsorption on γ-Fe₂O₃ at different initial diclofenac concentrations at 298 K (25 °C).

Three adsorption kinetic models, i.e., pseudo-first-order, pseudo-second-order and intraparticle diffusion, were used for modeling the kinetic data and consequently understanding the mechanism of diclofenac adsorption onto γ-Fe₂O₃ nanoparticles. The non-linear regression analysis of the experimental data using the pseudo-first- and pseudo-second-order models is shown in Fig. 3a while the linear regression analysis of the experimental data using the intraparticle diffusion model is presented in Fig. 3b. Fig. 3a shows that as the diclofenac concentration increases, the equilibrium is attained more slowly, i.e., 120 and 360 min for initial diclofenac concentrations of 100 and 500 mg L⁻¹, respectively. The diclofenac adsorption process was described in two stages, with an initial fast adsorption followed by a much slower adsorption rate afterward (Fig. 3b). The slower initial adsorption of diclofenac at the lower concentration suggests that the adsorption process occurs mainly on the γ-Fe₂O₃ adsorbent surface. This was expected since the pore distribution of the γ-Fe₂O₃ is in the range from 2 to 12 nm while the length of the diclofenac molecule is ∼1.08 nm, and at lower diclofenac concentration the driving force is smaller. Then, as the initial diclofenac concentration increases, its adsorption can be separated into two steps: (i) fast initial adsorption resulting from the fast attachment of the diclofenac molecules on the maghemite surface and (ii) slow adsorption resulting from the diffusion of the diclofenac molecules inside the pores.⁵³Table 2 shows the kinetic parameters estimated by fitting the kinetic data of diclofenac adsorption on maghemite with pseudo-first- and pseudo-second-order kinetic models while the kinetic parameters for the intraparticle diffusion model are given in Table 3. The kinetic data were better fitted by a pseudo-first-order model based on the χ_red² values for both initial concentrations of 100 and 500 mg L⁻¹. It is important to note that the experimental q_e values, i.e., q_e,exp,100 = 92.69 mg g⁻¹ and q_e,exp,500 = 240.50 mg g⁻¹, were also similar to the estimated q_e values obtained by the pseudo-first-order kinetic model. The parameter h, which indicates the initial rate of adsorption, decreased when the initial diclofenac concentration was increased.

Table 2 Kinetic parameters estimated using the pseudo-first- and pseudo-second-order models for diclofenac adsorption onto maghemite at 25 °C

C 0 (mg L^−1)	Pseudo-first-order
	q e,est (mg g^−1)	k 1 (min^−1)	h (mg g^−1 min^−1)	R ^2	χ red ^2
100	92.52	6.44 × 10^−2	5.96	0.9669	0.8217
500	249.94	8.98 × 10^−3	2.24	0.9864	69.3112

---

C 0 (mg L^−1)	Pseudo-second-order
	q e,est (mg g^−1)	k 2 (g mg^−1 min^−1)	h (mg g^−1 min^−1)	R ^2	χ red ^2
100	95.32	1.91 × 10^−3	17.36	0.9519	1.1949
500	332.62	2.37 × 10^−5	2.62	0.9810	96.6644

---

Table 3 Kinetic parameters estimated with the intraparticle diffusion model for diclofenac adsorption onto maghemite at 25 °C

C 0 (mg L^−1)	k i,1	k i,2	C i,1	C i,2	R i,1 ^2	R i,2 ^2
	(mg g^−1 min^−1)		(mg g^−1)
100	4.350	0.240	55.65	89.04	1	0.9623
500	23.024	9.189	–74.652	71.711	0.9977	0.8764

---

Different phenomena can control the adsorption process and determine its velocity. The external surface diffusion from the boundary layer and inside the pores were considered the most rate-limiting steps. Therefore, the kinetic results were also analyzed using the intraparticle diffusion model for a better understanding of the adsorption mechanism. The intraparticle diffusion plot, i.e., q_t against t^1/2, for the adsorption of diclofenac on the γ-Fe₂O₃ adsorbent is shown in Fig. 3b. Both plots presented multilinearity, suggesting at least two stages controlling the diclofenac adsorption process. The first stage indicates instantaneous adsorption, where the fastest stage is observed, and the second stage is a progressive adsorption where there is a limitation of the velocity due to the intraparticle diffusion. It is during the second stage that the adsorption equilibrium is attained after intraparticle diffusion begins to slow down probably due to three main reasons: (i) adsorption inside the pores; (ii) the electrostatic repulsion on the maghemite due to surface saturation; and (iii) the solute concentration in solution becomes very low, which decreased the driving force.⁵⁴

Table 3 shows the parameters estimated by fitting the intraparticle diffusion model to the kinetic data of diclofenac adsorption on the γ-Fe₂O₃ adsorbent. It can be noticed that the intraparticle diffusion rate constants, k_i,1, for the first adsorption step are higher than k_i,2 for both initial diclofenac concentrations, indicating that the adsorption rate was higher at the beginning and decreased over time.⁵⁵ Also, if the plot of q_t against t^1/2 yields a straight line that intersects the origin (C = 0), then pore diffusion is the only rate-limiting step controlling the adsorption of diclofenac on maghemite. However, the straight lines for both plots did not intersect the origin (C ≠ 0), and both plots were multilinear (Fig. 3b), suggesting that film diffusion or surface diffusion and pore diffusion are controlling the adsorption of diclofenac on the γ-Fe₂O₃.^56–58

Adsorption isotherms and thermodynamics of adsorption. To obtain the thermodynamic parameters of adsorption, i.e., changes in free energy of adsorption (Δ_adsG°), enthalpy of adsorption (Δ_adsH°) and entropy of adsorption (Δ_adsS°), adsorption isotherms of diclofenac on maghemite (pH 7.0) at five different temperatures from 283 K (10 °C) to 318 K (45 °C) were obtained. Fig. 4a shows the diclofenac adsorption isotherms at different temperatures fitted by the Langmuir model. The amount of diclofenac adsorbed increased with increasing temperature, suggesting that the adsorption process is endothermic. The adsorption capacity of diclofenac increased until 298 K, but no increase in the adsorption capacity was observed at higher temperatures, which may be due to saturation of the adsorbent sites. The equilibrium data obtained at different temperatures were fitted with the isotherm models of the Langmuir, Sips, Freundlich, and Temkin. The equilibrium data were better fitted to the Langmuir and Sips isotherm models, exhibiting few differences in χ_red² values. The Sips model is an isotherm model that can be used to evaluate the heterogeneity of the adsorption system. If the parameter n approaches unity, the Sips model can be reduced to the Langmuir model. The only value of parameter n that considerably deviated from unity was that obtained at 308 K (35 °C) (Table S1, ESI†). Therefore, the adsorption systems can be fitted by the Langmuir model, which does not necessarily mean that the adsorption systems studied followed the Langmuir model assumptions, since the surface of maghemite is not homogeneous and probably presents more than one type of adsorption site or adsorption sites with different affinities and energies, as pointed out by theoretical calculations.

Fig. 4 (a) Adsorption isotherms of diclofenac on the γ-Fe₂O₃ adsorbent at different temperatures at pH 7.0. (b) van’t Hoff plot for adsorption of diclofenac on the γ-Fe₂O₃ adsorbent at pH 7.0.

For a better understanding of the thermodynamics of the adsorption process studied, the Langmuir constants were adequately converted into equilibrium thermodynamic constants according to the approach reported by Liu³⁷ and Ghosal and Gupta³⁸ and used to obtain the changes in enthalpy and entropy of adsorption using the van’t Hoff plot (Fig. 4b). Then, the change in free energy of adsorption was calculated using the fundamental thermodynamic equation. The van’t Hoff plot exhibited a satisfactory fit to the experimental data with a coefficient of determination (R²) of 0.8669 and a χ_red² of 0.3399. Table 4 presents the thermodynamic parameters of adsorption. The change in enthalpy of adsorption indicates that the adsorption process of diclofenac on maghemite at pH 7.0 is endothermic, although there are many exothermic adsorption processes. However, the enthalpy of adsorption is a contribution of three subprocesses, i.e., desolvation of maghemite surface sites and diclofenac molecules, solute–solute interaction and interaction between γ-Fe₂O₃ surface sites and diclofenac molecules. Since the enthalpy of adsorption is positive, the energy released from interactions between maghemite surface sites and diclofenac molecules is lower than the energy involved in the desolvation of maghemite surface sites and diclofenac molecules and solute–solute interactions on the adsorbent surface.^59,60 On the other hand, interactions weaker than 84 kJ mol⁻¹ are typical of physisorption while the chemical bonding energies are between 84 and 420 kJ mol⁻¹.⁶¹ Thus, the adsorption process of diclofenac on maghemite can be classified as physisorption. Analyzing the value of the change in entropy of adsorption, it is comprehended that the negative value of the change in free energy of adsorption is a consequence of a thermodynamic compensation balance since the value of entropy of adsorption is much higher than the enthalpy of adsorption. Thus, the adsorption process of diclofenac on maghemite at pH 7.0 is entropically driven. The entropic contribution can be attributed to the release of water molecules from the maghemite surface caused by the migration of diclofenac molecules from bulk solution to the surface of the adsorbent, which presents oxygen atoms capable of making hydrogen bonds with water molecules.^59,60 Also, the free energy of adsorption becomes more negative as the temperature increases, thereby increasing the spontaneity of the adsorption process.

Table 4 Thermodynamic data for diclofenac adsorption on maghemite at pH 7.0

Temperature (K)	283	288	298^a	308	318
a Standard free energy of adsorption.
ΔadsG (kJ mol^−1)	−24.17	−25.78	−29.01	−32.24	−35.48
ΔadsS° (kJ mol^−1)			96.33
ΔadsH° (kJ mol^−1)			67.31

---

Adsorption studies as a function of pH and adsorption mechanism. For a better understanding of the adsorption mechanism as a function of pH, the adsorption isotherms of diclofenac were obtained at two different pH values. The point of zero charge (PZC) of maghemite was also determined. Fig. 5a shows a plot of ΔpH against pH_i used for the calculation of pH_PZC. The first derivative of the curve of ΔpH against pH_i (not shown) was used to estimate the pH_PZC, which was found to be 7.0. At a pH lower than pH_PZC, maghemite has a positive net surface charge, whereas at a pH higher than pH_PZC, it has a negative net surface charge. Fig. 5b shows the isotherms for the adsorption of diclofenac on maghemite at pH values of 7.0 and 10.0. The adsorption of diclofenac (pK_a = 4.0–4.15)⁶² by the γ-Fe₂O₃ adsorbent is favored at pH 7.0 in comparison to pH 10.0.⁶³ At pH 7.0, the net surface charge of maghemite is zero, assuming a pH_PZC of 7.0 for maghemite. In contrast, at a pH of 10.0, the surface of maghemite is negatively charged, increasing the repulsion between the surface of the γ-Fe₂O₃ adsorbent and negatively charged diclofenac molecules. Adsorption experiments using a more acidic pH, i.e., below 6, favored the formation of diclofenac flakes, possibly due to the pH approaching the pK_a of diclofenac, which varies between 4.0 and 4.15,⁶² thereby decreasing its solubility in water. As for most basic pH values above 10, the synthesized material reacts with hydroxyl ions and this changes the solution colour to yellow, preventing its use for water treatment.

Fig. 5 (a) ΔpH against pH_i curve for the determination of pH of the point of zero charge of maghemite. (b) Adsorption isotherms of diclofenac on maghemite at pH values of 7.0 and 10.0 at 298 K (25 °C) (Langmuir model parameters at pH = 10.0: Q_max = 137.42 mg g⁻¹; b = 0.078 L mg⁻¹; R² = 0.8993; χ_red² = 1083.78).

Fig. 6 shows a proposed mechanism for diclofenac adsorption onto the maghemite surface. The surface of the adsorbent (–M) has hydroxyl groups, which can probably adsorb diclofenac molecules by two main interactions, i.e., ion–dipole (Fig. 6a) and hydrogen bonding (Fig. 6b). In addition, an ion–dipole interaction between positive Fe atoms on the surface of the adsorbent and negatively charged oxygen atoms of the carboxylate groups can also occur. In the theoretical studies section, these modes of diclofenac attachment on the maghemite surface will be confirmed.

Fig. 6 Proposed mechanism for diclofenac adsorption onto the maghemite surface at pH 7.0.

To confirm the suggested mechanism for diclofenac adsorption on maghemite, a desorption study at different pH values was performed. The desorption percentage of diclofenac as a function of solution pH is shown in Fig. 7. The desorption percentage of diclofenac slightly increases as solution pH increases from 7 to 9, but the amount desorbed did not exceed 10%. However, at pH values higher than 9, the concentration of hydroxyl anions is increased, and these basic groups are probably responsible for deprotonating the hydroxyl groups on the maghemite surface. Therefore, at pH values higher than 9, the surface adsorption sites become negative (M–O⁻), which leads to the breakdown of the interactions between diclofenac molecules and the adsorption sites of maghemite. This promotes the desorption of diclofenac molecules from the surface of maghemite, thereby confirming the desorption mechanism proposed in Fig. 6c.

Fig. 7 Desorption of diclofenac from maghemite as a function of the solution pH at 298 K (25 °C).

Adsorption of diclofenac at low concentrations

Fig. S4a (see ESI†) shows the removal of diclofenac from aqueous solutions at a low concentration (5 mg L⁻¹) as a function of time, and Fig. S4b (ESI†) shows the adsorption isotherm of diclofenac at lower initial concentrations in comparison to those shown in Fig. 4a. The kinetic and equilibrium studies of diclofenac adsorption at lower initial concentrations showed that maghemite could be an efficient adsorbent for the treatment of surface waters contaminated with diclofenac.

Reuse of spent maghemite

The reuse of spent maghemite was performed for five cycles. Fig. S5 (see ESI†) shows the diclofenac removal efficiency as a function of the number of adsorption cycles. In the first two cycles, the diclofenac removal efficiency by maghemite is practically maintained; however, from the third cycle, a progressive decrease in the diclofenac removal efficiency by maghemite is observed, attaining ∼45% of its initial removal efficiency in the fifth cycle. Thus, although maghemite is an efficient adsorbent for diclofenac removal, more maghemite is needed in a water treatment plant (WTP) to ensure the complete removal of diclofenac as its efficiency decreases with its use along the adsorption cycles.

Theoretical studies. As previously mentioned, the XRD patterns shown in Fig. 1 are characteristic of maghemite. These values are in agreement with those reported by Cornell and U. Schwertmann,⁴⁹ who reported the following diffraction (hkl) planes (440), (511), (422), (400), (220), and (311), with the latter being the most intense. In line with that, theoretical calculations with the DFT method were used to study the maghemite surface before the adsorption of the diclofenac molecules. The results showed that the diffraction plane (311) was of lower energy, indicating higher stability. A DFT calculation was first carried out to examine the adsorption of the diclofenac molecule by approaching the diclofenac molecule onto the surface of maghemite in two ways, i.e., the diclofenac molecule is perpendicular (MP1 and MP4) or parallel (MP2 and MP3) to the maghemite surface (Fig. 8).

Fig. 8 Schematic view of diclofenac interacting with the maghemite surface in four arrangements: MP1: the functional group perpendicularly interacts with the surface; MP2: the functional group parallelly interacts with the surface; MP3: the functional group of DCF parallelly interacts with the surface; MP4: the functional group of DCF perpendicularly interacts with the surface.

Fig. 9 presents the plots of the energy values calculated for the different DCF/γ-Fe₂O₃ configurations. In this system, the perpendicular configurations, MP1 and MP4, yielded different energy values. In comparison to the perpendicular configurations, the energy values for both parallel configurations, MP2 and MP3, presented higher energy values. For the perpendicular and parallel (MP1, MP2, and MP3) adsorption models, the low energy values can be associated with physisorption. Two factors can be used to explain the change in the energy. One is the effective contact area between the diclofenac molecule and the maghemite surface, and the other is the atomic correlation between the DCF structural group and the maghemite surface. It is expected that the parallel configuration would be more favorable due to the interaction of the carboxylic group with the maghemite surface atoms. Analyzing the data reported in Fig. 9, the lowest energy value corresponds to the MP1 structure, with the carboxylic functional group interacting perpendicularly to the maghemite surface. This conformation was identified by a distance of 7 Å, which is consistent with the 700 pm distance between the diclofenac molecule and the maghemite surface.

Fig. 9 Evolution of energy values at different distances of the diclofenac molecule and the maghemite surface.

In a previous study, Rocha et al.⁵⁰ pointed out that the charge density is higher close to O atoms. This feature can lead to a significant electron transfer from Fe to O atoms in iron oxides. These results are consistent with the main features of the calculated density states map displayed in Fig. 10, which was found in the present study for the plane (311). The red areas around the iron atoms indicate the locations of the volumes with low electron density. In short, the electrostatic calculations carried out indicate the Fe as the sites for preferential bonding with the diclofenac molecule.

Fig. 10 Electrostatic surface contours of the maghemite surface, plane (311). Red and green indicate volumes of low electron density and high electron density, respectively.

The study of diclofenac adsorption as a function of solution pH showed that adsorption is favored at pH values closer to neutrality while for basic pH values the high basicity of the solution probably leads to the deprotonation of hydroxyl groups on the adsorbent surface and consequently to the repulsion between the negatively charged maghemite surface and diclofenac molecules. The structures used in the theoretical calculations simulate a neutral surface of maghemite with the approach of the deprotonated diclofenac molecule. The deprotonated O atoms are marked with an asterisk in Fig. 10.

Infrared spectroscopy analysis of maghemite loaded with DCF. Infrared (IR) spectroscopy was used for a better understanding of the mechanism of diclofenac adsorption on maghemite. Fig. S2 (ESI†) presents the infrared spectra for DCF, unloaded maghemite and maghemite loaded with DCF, respectively. In the IR spectrum of DCF, the two bands at 3391 and 3256 cm⁻¹, are attributed to the stretching vibrations of N–H of the secondary amine group bonded to the aromatic ring and O–H of the carboxylic group, respectively. The band at 1611 cm⁻¹ is related to the C=C stretching absorption of the aromatic ring. The bands at 1583 and 1396 cm⁻¹ correspond to the asymmetric and symmetric stretching of a carboxylate group, respectively, while the band at 1457 cm⁻¹ corresponds to CH₂ bending. The band at 1305 cm⁻¹ is attributed to the stretching vibration of the C–N bond while the band at 1047 cm⁻¹ is related to the C–Cl stretching in aryl chlorides.⁶⁴ New bands appeared in the IR spectrum of maghemite loaded with DCF when compared to the IR spectrum of unloaded maghemite, which is attributed to the adsorbed diclofenac molecules on the maghemite surface. The band at 3323 cm⁻¹ corresponds to the amine group of DCF. The shift of this band in the IR spectrum of DCF–maghemite when compared to the infrared spectrum of DCF may be related to hydrogen bond interactions occurring between the amine groups of DCF molecules and the hydroxyl groups in the maghemite surface. The spectrum of DCF–maghemite also exhibited two bands at 1589 and 1420 cm⁻¹, which are attributed to the stretching of the carboxylate group and are in good agreement with the theoretical calculations. These IR bands were shifted in the spectrum of DCF–maghemite in comparison with the spectrum of DCF. In addition, theoretical calculations showed that carboxylate groups would be the groups most favorable for interaction between diclofenac molecules and the maghemite surface. Furthermore, the vibrations of the Fe–O bonds were shifted from 563 cm⁻¹ to 552 cm⁻¹ and 582 cm⁻¹ due to the interactions with the diclofenac molecules.⁴⁷

Comparison with the literature data. The maximum adsorption capacity (Q_max) of maghemite is comparable to adsorption capacities for iron oxides reported in the literature (Table 5). The high diclofenac adsorption capacity was attributed to the small particle size and porosity as well as the functional groups present on the surface of the maghemite nanoparticles. Moreover, the mesoporous maghemite nanoparticles are magnetic, and their suspension in water can be easily separated by applying an external magnetic field. Considering a Q_max value of ∼250 mg g⁻¹, a DCF concentration of 400 ng L⁻¹ in Brazilian surface waters⁷³ and that 90% is to be removed in a water treatment plant (WTP), the estimated dose of maghemite adsorbent would be 1.5 × 10⁻⁶ g L⁻¹, indicating that 1.5 g of maghemite would be enough to treat 1000 m³ of water, or that 150 g d⁻¹ of the material would be needed to treat water that feeds 50 000 inhabitants (considering a per capita water consumption of 200 L hab⁻¹ d⁻¹). However, since there might be other contaminants competing with DCF for the adsorption sites of maghemite, a higher amount would probably be required. In this case, adsorption tests with environmental samples (multicomponent matrix) are needed to ascertain the amount of maghemite needed for a given treatment. It should be highlighted that this high adsorption capacity was found for the unmodified maghemite adsorbent, and therefore, several possibilities to improve the adsorption capacity of this magnetic iron oxide based on surface chemical modification are still open for further studies.

Table 5 Comparison of the maximum adsorption capacities for diclofenac of different adsorbents

Adsorbent	Q max (mg g^−1) at 298 K	pH	Ref.
Reduced graphene oxide	59.67	10.0	65
Montmorillonite	497.00	6.7	66
Chloride-modified zeolite	31.35	7.4	67
Hexadecyltrimethylammonium bromide-modified zeolite	45.87	7.4	67
Modified diatomite	390.02	7.5	68
Graphene oxide	500.00	7.0	69
Isabel grape bagasse	76.98	5.0	70
Molecularly imprinted polymer	324.80	8.0	71
Granular carbon nanotubes/alumina hybrid	33.86	6.0	72
Commercial activated carbon	83	5.5	70
This study	120.00	10.0
This study	252.00	7.0	—

---

Conclusions

Maghemite was synthesized by a new synthesis route using low-cost and non-toxic reagents. The synthesized maghemite had a high surface area and a small particle size, as suggested by the data obtained from BET/BJH and XRD. The maximum adsorption capacity for diclofenac in an aqueous solution was 252 mg g⁻¹ at pH 7.0. Thermodynamic studies indicated that the adsorption of diclofenac on maghemite is an endothermic process. The Langmuir and Sips models best fitted the equilibrium data obtained while the adsorption kinetics followed a pseudo-first-order model. Adsorption experiments at different pH values showed that the solution pH was a crucial parameter for diclofenac removal by maghemite nanoparticles. At pH higher than 9, diclofenac removal was hindered, possibly due to its competition with hydroxyl ions for the active sites of the material. The theoretical calculations reinforce the adsorption mechanism proposed (physisorption), in which the diclofenac molecules bind on the maghemite surface by using OH functional groups.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors are grateful to the Federal University of Ouro Preto (UFOP grant number 23109.003209/2016-98), the Brazilian National Research Council (CNPq grant number 443426/2014-7), the Minas Gerais State Research Funding Agency (FAPEMIG grant number APQ-00847-14; APQ-00369-14) and the National Council for the Improvement of Higher Education (CAPES) for funding this research.

Notes and references

1. B. Petrie, R. Barden and B. Kasprzyk-Hordern, Water Res., 2015, 72, 3–27.
2. C. Jung, A. Son, N. Her, K.-D. Zoh, J. Cho and Y. Yoon, J. Ind. Eng. Chem., 2015, 27, 1–11.
3. European Parliament, Off. J. Eur. Communities: Legis., 2008, 164, 19–40.
4. S. Álvarez-Torrellas, A. Rodríguez, G. Ovejero and J. García, Chem. Eng. J., 2016, 283, 936–947.
5. A. Ziylan and N. H. Ince, J. Hazard. Mater., 2011, 187, 24–36.
6. B. t. Ferrari, N. Paxéus, R. L. Giudice, A. Pollio and J. Garric, Ecotoxicol. Environ. Saf., 2003, 55, 359–370.
7. D. G. Oviedo-Gomez, M. Galar-Martinez, S. Garcia-Medina, C. Razo-Estrada and L. M. Gomez-Olivan, Environ. Toxicol. Pharmacol., 2010, 29, 39–43.
8. A. G. Slater and A. I. Cooper, Science, 2015, 348, aaa8075.
9. R. E. Morris and J. Cejka, Nat. Chem., 2015, 7, 381–388.
10. B. Ashourirad, A. K. Sekizkardes, S. Altarawneh and H. M. El-Kaderi, Chem. Mater., 2015, 27, 1349–1358.
11. N. Chalal, H. Bouhali, H. Hamaizi, B. Lebeau and A. Bengueddach, Microporous Mesoporous Mater., 2015, 210, 32–38.
12. H. Zhao, X. Liu, Z. Cao, Y. Zhan, X. Shi, Y. Yang, J. Zhou and J. Xu, J. Hazard. Mater., 2016, 310, 235–245.
13. B. N. Bhadra, P. W. Seo and S. H. Jhung, Chem. Eng. J., 2016, 301, 27–34.
14. M. J. Ahmed, J. Environ. Manage., 2017, 190, 274–282.
15. H. L. Derang Cao, L. Pan, J. Li, X. Wang, P. Jing, X. Cheng, W. Wang, J. Wang and Q. Liu, Sci. Rep., 2016, 6, 32360.
16. J. Lalley, C. Han, X. Li, D. D. Dionysiou and M. N. Nadagouda, Chem. Eng. J., 2016, 284, 1386–1396.
17. N. V. Long, Y. Yang, T. Teranishi, C. M. Thi, Y. Cao and M. Nogami, Mater. Des., 2015, 86, 797–808.
18. M. C. Pereira, L. C. A. Oliveira and E. Murad, Clay Miner., 2012, 47, 285–302.
19. O. Horner, S. Neveu, S. de Montredon, J.-M. Siaugue and V. Cabuil, J. Nanopart. Res., 2009, 11, 1247–1250.
20. J. Lu, X. Jiao, D. Chen and W. Li, J. Phys. Chem. C, 2009, 113, 4012–4017.
21. J. Vidal-Vidal, J. Rivas and M. López-Quintela, Colloids Surf., A, 2006, 288, 44–51.
22. M. Aliahmad and N. Nasiri Moghaddam, Mater. Sci.-Pol., 2013, 31, 264–268.
23. P. Scherrer, Nachr. Ges. Wiss. Goettingen, Math.-Phys. Kl., 1918, 98–100.
24. S. Brunauer, P. H. Emmett and E. Teller, J. Am. Chem. Soc., 1938, 60, 309–319.
25. E. P. Barrett, L. G. Joyner and P. P. Halenda, J. Am. Chem. Soc., 1951, 73, 373–380.
26. L. Ai, C. Zhang, F. Liao, Y. Wang, M. Li, L. Meng and J. Jiang, J. Hazard. Mater., 2011, 198, 282–290.
27. S. Y. Lagergren, Handlingar, 1898, 24, 1–39.
28. Y. S. Ho and G. McKay, Process Biochem., 1999, 34, 451–465.
29. T. W. Weber and R. K. Chakravorti, AIChE J., 1974, 20, 228–238.
30. T. A. Khan, S. A. Chaudhry and I. Ali, J. Mol. Liq., 2015, 202, 165–175.
31. I. Langmuir, J. Am. Chem. Soc., 1918, 40, 1361–1403.
32. H. M. Freundlich, Z. Phys. Chem., 1906, 57, 385–470.
33. M. Temkin and V. Pyzhev, Acta Physicochim. URSS, 1940, 12, 217–222.
34. R. Sips, J. Chem. Phys., 1948, 16, 490–495.
35. O. Hamdaoui and E. Naffrechoux, J. Hazard. Mater., 2007, 147, 381–394.
36. J. H. Van't Hoff, Etudes de dynamique chimique, Frederik Mullet and Co., Amsterdam, 1884.
37. Y. Liu, J. Chem. Eng. Data, 2009, 54, 1981–1985.
38. P. S. Ghosal and A. K. Gupta, J. Mol. Liq., 2017, 225, 137–146.
39. E. Guntelberg, Z. Phys. Chem., 1926, 123, 199–247.
40. G. te Velde, F. M. Bickelhaupt, E. J. Baerends, C. Fonseca Guerra, S. J. A. van Gisbergen, J. G. Snijders and T. Ziegler, J. Comput. Chem., 2001, 22, 931–967.
41. E. J. B. G. te Velde, P. H. T. Philipsen, G. Wiesenekker, J. A. Groeneveld, J. A. Berger, P. L. de Boeij and R. Klooster, F. Kootstra, P. Romaniello, J. G. Snijders, E. S. Kadantsev, T. Ziegler, Vrije Universiteit, Amsterdam, The Netherlands, 2009.
42. Q. Yang, D. Liu, C. Zhong and J.-R. Li, Chem. Rev., 2013, 113, 8261–8323.
43. E. Darezereshki, M. Ranjbar and F. Bakhtiari, J. Alloys Compd., 2010, 502, 257–260.
44. W. Jiang, M. Pelaez, D. D. Dionysiou, M. H. Entezari, D. Tsoutsou and K. O’Shea, Chem. Eng. J., 2013, 222, 527–533.
45. A. Afkhami and R. Moosavi, J. Hazard. Mater., 2010, 174, 398–403.
46. J. Hu, G. Chen and I. M. C. Lo, J. Environ. Eng., 2006, 132, 709–715.
47. S. Rajput, L. P. Singh, C. U. Pittman and D. Mohan, J. Colloid Interface Sci., 2017, 492, 176–190.
48. A. Etale, H. Tutu and D. C. Drake, Appl. Water Sci., 2016, 6, 187–197.
49. R. M. Cornell and U. Schwertmann, The Iron Oxides: Structure, Properties, Reactions, Occurrences and Uses, Wiley-VCH, Darmstadt, Germany, 2nd edn, 2006.
50. M. V. J. Rocha, H. W. P. de Carvalho, V. H. V. Sarmento, A. F. Craievich and T. C. Ramalho, Polym. Compos., 2016, 37, 51–60.
51. F. Rojas, I. Kornhauser, C. Felipe, J. M. Esparza, S. Cordero, A. Dominguez and J. L. Riccardo, Phys. Chem. Chem. Phys., 2002, 4, 2346–2355.
52. G. Corro, F. Bañuelos, E. Vidal and S. Cebada, Fuel, 2014, 115, 625–628.
53. D. Lázaro, C. Caldeira, M. Dantas, M. Mansur, L. Oliveira and A. Franca, Int. J. Sustain. Dev. Plann., 2012, 7, 446–456.
54. W. J. Weber and E. H. Smith, Environ. Sci. Technol., 1987, 21, 1040–1050.
55. L. Largitte and R. Pasquier, Chem. Eng. Res. Des., 2016, 109, 495–504.
56. M. A. Bahri, L. Calvo, J. Lemus, M. A. Gilarranz, J. Palomar and J. J. Rodriguez, Chem. Eng. J., 2012, 198–199, 346–354.
57. K. A. G. Gusmão, L. V. A. Gurgel, T. M. S. Melo and L. F. Gil, J. Environ. Manage., 2013, 118, 135–143.
58. L. A. Al-Khateeb, A. Y. Obaid, N. A. Asiri and M. Abdel Salam, J. Ind. Eng. Chem., 2014, 20, 916–924.
59. F. S. Teodoro, S. N. D. C. Ramos, M. M. C. Elias, A. B. Mageste, G. M. D. Ferreira, L. H. M. da Silva, L. F. Gil and L. V. A. Gurgel, J. Colloid Interface Sci., 2016, 483, 185–200.
60. B. C. S. Ferreira, F. S. Teodoro, A. B. Mageste, L. F. Gil, R. P. de Freitas and L. V. A. Gurgel, Ind. Crops Prod., 2015, 65C, 521–534.
61. H. Mahmoodian, O. Moradi, B. Shariatzadeha, T. A. Salehf, I. Tyagi, A. Maity, M. Asif and V. K. Gupta, J. Mol. Liq., 2015, 202, 189–198.
62. H. Zhao, X. Liu, Z. Cao, Y. Zhan, X. Shi, Y. Yang, J. Zhou and J. Xu, J. Hazard. Mater., 2016, 310, 235–245.
63. G. Lamanna, M. Kueny-Stotz, H. Mamlouk-Chaouachi, C. Ghobril, B. Basly, A. Bertin, I. Miladi, C. Billotey, G. Pourroy, S. Begin-Colin and D. Felder-Flesch, Biomaterials, 2011, 32, 8562–8573.
64. D. Pavia, G. Lampman, G. Kriz and J. Vyvyan, Introduction to Spectroscopy, Cengage Learning, 4th edn, 2014.
65. I. M. Jauris, C. F. Matos, C. Saucier, E. C. Lima, A. J. G. Zarbin, S. B. Fagan, F. M. Machado and I. Zanella, Phys. Chem. Chem. Phys., 2016, 18, 1526–1536.
66. M. Kaur and M. Datta, Adsorpt. Sci. Technol., 2014, 32, 365–387.
67. D. Krajisnik, A. Dakovic, M. Milojevic, A. Malenovic, M. Kragovic, D. B. Bogdanovic, V. Dondur and J. Milic, Colloids Surf., B, 2011, 83, 165–172.
68. J. Janicijevic, D. Krajisnik, B. Calija, B. N. Vasiljevic, V. Dobricic, A. Dakovic, M. D. Antonijevic and J. Milic, Int. J. Pharm., 2015, 496, 466–474.
69. S.-W. Nam, C. Jung, H. Li, M. Yu, J. R. V. Flora, L. K. Boateng, N. Her, K.-D. Zoh and Y. Yoon, Chemosphere, 2015, 136, 20–26.
70. M. Antunes, V. I. Esteves, R. Guégan, J. S. Crespo, A. N. Fernandes and M. Giovanela, Chem. Eng. J., 2012, 192, 114–121.
71. C. M. Dai, S. U. Geissen, Y. L. Zhang, Y. J. Zhang and X. F. Zhou, Environ. Pollut., 2011, 159, 1660–1666.
72. H. Wei, S. Deng, Q. Huang, Y. Nie, B. Wang, J. Huang and G. Yu, Water Res., 2013, 47, 4139–4147.
73. D. R. S. Lima, M. C. Tonucci, S. F. Aquino and M. Libânio, Eng. Sanit. Ambiental, 2017 , in press.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7nj03214e

---