Novel Magnetic Nanocomposites Based on Carboxyl-Functionalized SBA-15 Silica for Effective Dye Adsorption from Aqueous Solutions

In this study, three novel magnetic nanocomposites based on carboxyl-functionalized SBA-15 silica and magnetite nanoparticles were prepared through an effective and simple procedure and applied for methylene blue (MB) and malachite green G (MG) adsorption from single and binary solutions. Structure, composition, morphology, magnetic, and textural properties of the composites were thoroughly investigated. The influence of the amount of carboxyl functional groups on the physicochemical and adsorptive properties of the final materials was investigated. The capacity of the synthesized composites to adsorb MB and MG from single and binary solutions and the factors affecting the adsorption process, such as contact time, solution pH, and dye concentration, were assessed. Kinetic modelling showed that the dye adsorption mechanism followed the pseudo-second-order kinetic model, indicating that adsorption was a chemically controlled multilayer process. The adsorption rate was simultaneously controlled by external film diffusion and intraparticle diffusion. It was evidenced that the molecular geometry of the dye molecule plays a major role in the adsorption process, with the planar geometry of the MB molecule favoring adsorption. The analysis of equilibrium data revealed the best description of MB adsorption behavior by the Langmuir isotherm model, whereas the Freundlich model described better the MG adsorption.


Introduction
One of the most serious environmental problems in modern society is water pollution, mainly due to effluents from textile, leather, paints and pigments, plastics, paper, food, and cosmetic industries [1]. Among the different pollutants, dyes represent one of the most problematic groups of organic pollutants due to their toxicity, which causes serious health effects on animals and human beings. Most of them are stable to biodegradation, photodegradation, and oxidizing agents. Some studies revealed their carcinogenic effects due to the changes induced in DNA synthesis [2,3]. The aquatic organisms are also affected by the presence of dyes in water because they hinder light penetration through the water surface decreasing the photosynthetic activity of phytoplankton. This causes oxygen deficiency and disturbs the biological cycle of aquatic biota [4]. Moreover, dye molecules in wastewater lead to mutagenicity, carcinogenicity, and the dysfunction of human beings' kidney, liver, brain, reproductive system, and central nervous system [5].
The aim of this paper was to obtain magnetic nanocomposites based on carboxylfunctionalized SBA-15 silica and magnetite nanoparticles through a convenient and effective procedure and to evaluate their effectiveness as adsorbents for cationic dyes. The influence of the amount of carboxyl functional groups on the physicochemical and adsorptive properties of the final nanomaterials was investigated. All the nanocomposites were characterized by nitrogen adsorption-desorption analysis, Fourier-transform infrared (FT-IR) spectroscopy, thermogravimetric (TGA) and elemental analysis, zeta potential, Xray diffraction (XRD), scanning/transmission electron microscopy (SEM/TEM), and the vibrating-sample magnetometer technique (VSM). Methylene blue (MB) and malachite green G (MG) (Figure 1) were used as model pollutants in the adsorption tests. The adsorption process was investigated as a function of solution pH, contact time, and initial concentration of dye. The characteristics of the adsorption isotherms and kinetics were also studied in single and binary solutions.

Synthesis of Magnetic Nanocomposites Based on Carboxyl-Functionalized SBA-15 Silica
3.0 g of P-123 was dissolved in 93 mL hydrochloric acid 2M solution until the solution became clear. 2-Cyanoethyltriethoxysilane was then added and the solution was stirred at 40 °C for 30 min. The mixture was transferred to a polypropylene bottle, then TEOS was slowly added and the mixture was kept under stirring at 40 °C for 22 h, followed by aging in an oven at 100 °C for 45 h, under static conditions. The solid product was separated by centrifugation, washed several times with distilled water and then dried at 90 °C in oven.
The molar composition of the mixture was (1 − x) TEOS: x CTES: 6 HCl: 167 H2O: 0.017 P123, where x = 0, 0.05, and 0.1. For the hydrolysis of the -CN groups, the product was treated with 150 mL 48 wt. % H2SO4 at 95 °C for 24 h [21,22]. Subsequently, the mixture was filtered on a G4 fritted filter funnel and the solid product was washed with a

Synthesis of Magnetic Nanocomposites Based on Carboxyl-Functionalized SBA-15 Silica
3.0 g of P-123 was dissolved in 93 mL hydrochloric acid 2M solution until the solution became clear. 2-Cyanoethyltriethoxysilane was then added and the solution was stirred at 40 • C for 30 min. The mixture was transferred to a polypropylene bottle, then TEOS was slowly added and the mixture was kept under stirring at 40 • C for 22 h, followed by aging in an oven at 100 • C for 45 h, under static conditions. The solid product was separated by centrifugation, washed several times with distilled water and then dried at 90 • C in oven.
The molar composition of the mixture was (1 − x) TEOS: x CTES: 6 HCl: 167 H 2 O: 0.017 P123, where x = 0, 0.05, and 0.1. For the hydrolysis of the -CN groups, the product was treated with 150 mL 48 wt.% H 2 SO 4 at 95 • C for 24 h [21,22]. Subsequently, the mixture was filtered on a G4 fritted filter funnel and the solid product was washed with a copious amount of water until the eluent became neutral, then several times with hot ethanol. The obtained carboxyl-functionalized mesoporous silica samples were denoted: SBA15-COOH-x, where x = 0, 0.05, and 0.1. The magnetite particles were prepared according to a previous paper [23]. To an aqueous solution of 2.77 mmol FeCl 3 ·6H 2 O and 1.38 mmol FeCl 2 ·4H 2 O in 60 mL deionized water, 30 mL of 25% NH 4 OH solution was added under vigorous stirring. After heating at 80 • C for one hour, the resulting black precipitate was separated with the aid of a neodymium magnet and washed with distilled water until the pH of the runoff was neutral. The as-prepared magnetite particles were dispersed in water/ethanol mixture (volume ratio 4:3) by sonication; then, 1.2 mL of 25% NH 4 OH solution and 0.3 mL TEOS were consecutively added to the suspension; then, the mixture was sonicated 20 min. To this mixture, a certain amount of SBA15-COOHx dispersed in water/ethanol (volume ratio 4:3) was added under strong stirring. The mixture was sonicated for 45 min, then stirred continuously for 20 h at room temperature. The obtained solid materials were separated by centrifugation, washed three times with water, and dried in air. The final magnetic nanocomposites were denoted Fe 3 O 4 -SBA15, Fe 3 O 4 -SBA15-COOH-0.05, and Fe 3 O 4 -SBA15-COOH-0.1.

Characterization Methods
The nitrogen adsorption-desorption isotherms at −196 • C were measured using a Micromeritics ASAP 2020 analyzer (Norcross, GA, USA). The samples were degassed at 100 • C for 6 h under vacuum before analysis. Specific surface areas (S BET ) were estimated according to Brunauer-Emmett-Teller (BET) method, using adsorption data in the relative pressure range 0.05-0.30. The total pore volume (V total ) was calculated from the amount adsorbed at the relative pressure of 0.99. The average pore diameter and pore size distribution curves were obtained using Barrett-Joyner-Halenda (BJH) method using the desorption branch. FT-IR spectra were obtained on a Jasco FT/IR-4700 spectrophotometer (Tokyo, Japan) using KBr pellets. Thermogravimetric analyses (TGA) were performed using a Mettler Toledo TGA/SDTA851e thermogravimeter (Mettler Toledo, Greifensee, Switzerland), using a heating rate of 10 • C min −1 , open alumina crucibles, and a 20 mL min −1 synthetic air flow. The content of C, H, and N was determined through elemental analysis on a EuroEA elemental analyzer (HEKAtech GmbH, Wegberg, Germany). Zeta potential was measured by electrophoretic light scattering using a Backman Coulter Delsa Nano C particle analyzer (Brea, CA, USA) with 100 µg/mL sample dispersed in distilled water. Samples were ultrasonicated for 15 min before measurement. The powder X-Ray diffraction (XRD) analysis was performed on a Rigaku Ultima IV diffractometer (Rigaku Co., Tokyo, Japan) using a monochromatic Cu Kα (λ = 1.5418 Å) radiation source operated at 40 kV and 30 mA. The wide-angle diffractograms were recorded in the 2θ range 10-80 • , with 2 • min −1 scan speed and 0.02 • step width, while the low-angle diffractograms were recorded between 0.6 and 5 • , with 1 • min −1 scan speed and 0.02 • step width. XRD data were analyzed using Rigaku's PDXL software connected to ICDD PDF-2 database. The average crystallite size was calculated using the Williamson-Hall method. Microstructural analysis was performed on a TESCAN LYRA 3 XMU scanning electron microscope (SEM) (Tescan Orsay Holding, Brno-Kohoutovice, Czech Republic). JEOL 2100 electron microscope equipped with LaB6 filament and high-resolution polar piece (JEOL GmbH, Freising, Germany) was used for transmission electron microscopy (TEM) investigations. The magnetic properties were measured at room temperature on a Lake Shore's fully integrated Vibrating-Sample Magnetometer system 7404 (VSM) (Westerville, OH, USA). The experimental data were analyzed by fitting to the Langevin function.

Adsorption Experiments
The effect of pH, time, and concentration of dye solution on the adsorption capacity of the samples was studied via batch experiments in 100 mL conical flasks containing 5 mg adsorbent/20 mL of each of the dye solutions. All mixtures were stirred at 150 rpm on a GFL 3015 orbital shaker (Burgwedel, Germany). Stock solutions of dyes were obtained by dissolving 1 g of the respective dye in 1 L of distilled water, followed by dilution to obtain the desired concentration: between 5 and 100 mg/L. HCl and NH 4 OH solutions of various concentrations were used for dye solution's pH changing. The pH of the dye solution was determined at room temperature using an Agilent 3200 laboratory pH meter (Agilent Technologies, Shanghai, China). After achieving equilibrium, the dyeloaded adsorbents were separated using a hand-held magnet and the dye concentration in the remaining solutions was determined. MB and MG analysis was performed on an Agilent 1200 series HPLC (Tokyo, Japan) system equipped with: semipermeable membrane degasser, quaternary pump, autosampler with variable injection volume (0.1-100 µL), thermostatted column compartment, and a diode array detector (DAD) with the ability to record simultaneously UV-VIS spectra (190-900 nm) and up to 8 discrete wavelengths in this range. All chromatographic runs were carried out on an Acclaim Surfactant Plus column (150 × 3.0 mm, 3.0 µm) from Thermo Scientific. The detection of the target compounds was performed at the absorption maximum of λ = 665 nm (MB) and λ = 610 nm (MG) which were observed in the UV-Vis spectrum obtained by HPLC-DAD. Agilent ChemStation software was used for data acquisition, processing, and reporting.
The adsorption capacity was calculated using the following formula: where Q e -the amount of dye adsorbed at equilibrium (mg g −1 ), C i -the initial concentration of dye solution (mg L −1 ), C e -the equilibrium concentration of dye solution (mg L −1 ), V-total volume of dye solution (L), and m-mass of adsorbent used (g). All the adsorption tests were performed in triplicate with a maximum experimental error of 5%.

Materials Characterization
The FTIR spectra of magnetite-containing samples are shown in Figure 2. The band corresponding to the stretching vibration of the C=O bond of carboxyl groups, at 1721 cm −1 , is clearly evidenced in the spectra of Fe 3 O 4 -SBA15-COOH-0.05 and Fe 3 O 4 -SBA15-COOH-0.1 and is absent in the spectrum of Fe 3 O 4 -SBA15. The appearance of this band and the absence of that corresponding to CN-stretching vibration (2252 cm −1 ) confirms the hydrolysis of all cyano groups [21,22]. A close inspection of this band revealed that its intensity increases with the content of CTES in the silica source. This confirms that the amount of carboxyl groups in Fe 3 O 4 -SBA15-COOH-0.1 is higher than in Fe 3 O 4 -SBA15-COOH-0.05. This aspect is much more evident in the spectra of carboxyl-functionalized SBA-15 silica samples without magnetite content ( Figure S1). Moreover, a slight displacement of this band from 1717 to 1721 cm −1 is observed in the case of magnetic nanocomposites compared to the corresponding samples of carboxyl-functionalized silica. This shift could be attributed to the interactions of carboxyl groups with the surface of magnetite nanoparticles. The bands at 1084 and 800 cm −1 correspond to asymmetric and symmetric stretching vibration of Si-O-Si, while that at 465 cm −1 to asymmetric deformation vibration of O-Si-O [24]. The stretching vibrations of hydroxyl groups on the silica surface and those of adsorbed water appear at 1632 and 3435 cm −1 . The bands at 2923 and 2856 cm −1 are assigned to asymmetric and symmetric vibrations of C-H bonds in the ethyl chain anchored on the silica surface. The presence of magnetite nanoparticles was confirmed by the absorption band at 580 cm −1 which corresponds to the Fe-O bond [25]. Figure 3 shows the nitrogen adsorption-desorption isotherms and pore size distribution curves of the magnetic nanocomposites. All these samples exhibit type IV isotherms with an H1 type hysteresis characteristic for materials with ordered mesoporous structure, according to IUPAC classification [26]. One can notice that the CTES molar content in the total silica source has a deep effect on the formation of SBA-15 ordered mesostructure. As the CTES content increases, the hysteresis closure point shifts slightly to lower p/p • values, while the desorption branch becomes less steep, indicating a decrease in average pore diameters. This observation is also confirmed by the pore size distribution (PSD) calculated by the BJH method ( Figure 3, right side). The increasing of CTES content also leads to the decrease in specific surface area of the carboxylated-SBA-15 samples compared with pure SBA15, which proves the immobilization of carboxyl functional groups on the internal surface of mesoporous silica channels (Table 1). Judging from the shape of the hysteresis loop, it is clear that the introduction of a larger amount of CTES into the synthesis will lead to the destruction of the uniform mesoporous structure. The magnetic nanocomposites have lower surface areas and pore volumes than carboxylated-SBA15 samples, but close to each other, while the PSD curves are similar to those of the corresponding carboxylated-SBA15 materials ( Figure S2). This is an indication that the mesoporous silica interacted with the magnetite nanoparticles mainly on the external surface through the carboxyl groups which facilitated the interaction and to a lesser extent at the entrance of the pores.  Figure 3 shows the nitrogen adsorption-desorption isotherms and pore size distribution curves of the magnetic nanocomposites. All these samples exhibit type IV isotherms with an H1 type hysteresis characteristic for materials with ordered mesoporous structure, according to IUPAC classification [26]. One can notice that the CTES molar content in the total silica source has a deep effect on the formation of SBA-15 ordered mesostructure. As the CTES content increases, the hysteresis closure point shifts slightly to lower p/p0 values, while the desorption branch becomes less steep, indicating a decrease in average pore diameters. This observation is also confirmed by the pore size distribution (PSD) calculated by the BJH method ( Figure 3, right side). The increasing of CTES content also leads to the decrease in specific surface area of the carboxylated-SBA-15 samples compared with pure SBA15, which proves the immobilization of carboxyl functional groups on the internal surface of mesoporous silica channels (Table 1). Judging from the shape of the hysteresis loop, it is clear that the introduction of a larger amount of CTES into the synthesis will lead to the destruction of the uniform mesoporous structure. The magnetic nanocomposites have lower surface areas and pore volumes than carboxylated-SBA15 samples, but close to each other, while the PSD curves are similar to those of the corresponding carboxylated-SBA15 materials ( Figure S2). This is an indication that the mesoporous silica interacted with the magnetite nanoparticles mainly on the external surface through the carboxyl groups which facilitated the interaction and to a lesser extent at the entrance of the pores.   Figure 3 shows the nitrogen adsorption-desorption isotherms and pore size distribution curves of the magnetic nanocomposites. All these samples exhibit type IV isotherms with an H1 type hysteresis characteristic for materials with ordered mesoporous structure, according to IUPAC classification [26]. One can notice that the CTES molar content in the total silica source has a deep effect on the formation of SBA-15 ordered mesostructure. As the CTES content increases, the hysteresis closure point shifts slightly to lower p/p0 values, while the desorption branch becomes less steep, indicating a decrease in average pore diameters. This observation is also confirmed by the pore size distribution (PSD) calculated by the BJH method ( Figure 3, right side). The increasing of CTES content also leads to the decrease in specific surface area of the carboxylated-SBA-15 samples compared with pure SBA15, which proves the immobilization of carboxyl functional groups on the internal surface of mesoporous silica channels (Table 1). Judging from the shape of the hysteresis loop, it is clear that the introduction of a larger amount of CTES into the synthesis will lead to the destruction of the uniform mesoporous structure. The magnetic nanocomposites have lower surface areas and pore volumes than carboxylated-SBA15 samples, but close to each other, while the PSD curves are similar to those of the corresponding carboxylated-SBA15 materials ( Figure S2). This is an indication that the mesoporous silica interacted with the magnetite nanoparticles mainly on the external surface through the carboxyl groups which facilitated the interaction and to a lesser extent at the entrance of the pores.   from the relatively ordered structure of the SBA15. Figure 4b shows the wide-angle XRD powder diffraction patterns of the samples. The reflections at 2theta: 29.98(220), 35.58(311), 43.05(400), 53.48(442), 57.14(511), 62.92(440), and 74.30(533) values, confirm the presence of magnetite with an inverse cubic spinel structure in the investigated samples [28]. The crystallite size of the magnetite was calculated to be~6 nm.  Figure 4a shows the low-angle XRD powder diffraction patterns of the samples. Three well-resolved peaks which can be indexed as (100), (110), and (200) diffraction peaks associated with P6mm hexagonal symmetry of SBA-15 can be observed for Fe3O4-SBA15 and Fe3O4-SBA15-COOH-0.05 [27]. This suggests a highly ordered hexagonal structure of the SBA15 silica in these two samples. For the sample Fe3O4-SBA15-COOH-0.1, the intensity of the diffraction peaks decreased compared with the other two samples, indicating that the increasing of the content of COOH functional groups induces a deviation from the relatively ordered structure of the SBA15. Figure 4b shows the wide-angle XRD powder diffraction patterns of the samples. The reflections at 2theta: 29.98(220), 35.58(311), 43.05(400), 53.48(442), 57.14(511), 62.92(440), and 74.30(533) values, confirm the presence of magnetite with an inverse cubic spinel structure in the investigated samples [28]. The crystallite size of the magnetite was calculated to be ~6 nm.   (Figure 5c). However, the worm-like SBA-15 particles included in the magnetic composites have shorter one-dimensional pore channels which could facilitate mass diffusion within the pore channels.
Transmission electron microscopy (TEM) images of the magnetic composites and bare SBA-15 are shown in Figure 6. All TEM images clearly display the parallel arrangement of hexagonal pore channels of~8 nm, characteristic of SBA-15 silica. It can be noticed that the highly ordered mesoporous structure of SBA-15 was preserved regardless of the COOH content. The dark small spots of~8 nm represent the magnetite nanoparticles, randomly distributed on the SBA-15 surface. The elemental mapping shown in Figure 7   have shorter one-dimensional pore channels which could facilitate mass diffusion within the pore channels. Transmission electron microscopy (TEM) images of the magnetic composites and bare SBA-15 are shown in Figure 6. All TEM images clearly display the parallel arrangement of hexagonal pore channels of ~ 8 nm, characteristic of SBA-15 silica. It can be noticed that the highly ordered mesoporous structure of SBA-15 was preserved regardless of the COOH content. The dark small spots of ~8 nm represent the magnetite nanoparticles, randomly distributed on the SBA-15 surface. The elemental mapping shown in Figure 7 only for Fe3O4-SBA15 shows the homogeneous distribution of Fe3O4 nanoparticles onto the surface of SBA-15. Thermogravimetric analyses (TGA) were carried out in order to investigate the thermal behavior and composition of the magnetic nanocomposites. The loss of physisorbed water can be noticed for all materials on heating up from 25 to 140 • C ( Figure 8). Two superimposed mass loss events can be noticed in the temperature range 140-650 • C for all the samples. These two thermal events roughly centered at 300 and 500 • C correspond to the combustion of organic groups and the oxidation of magnetite to Fe 2 O 3 . The oxidation of Fe 3 O 4 to Fe 2 O 3 after the heat treatment was confirmed visually, as the materials changed color from brown-black before the TG analysis to the characteristic red color of Fe 2 O 3 . The mass decrease up to 650 • C corresponds to 4.1, 6.6, and 8.5% wt. for the Fe 3 O 4 -SBA15, Fe 3 O 4 -SBA15-COOH-0.05, and Fe 3 O 4 -SBA15-COOH-0.1 samples, respectively. All three samples exhibit a gradual, 1.1% mass loss above 650 • C, which can be ascribed to the dehydration of silanol groups [29]. The percent of organic material was calculated by subtracting the data obtained by TGA for the carboxyl-containing samples and Fe 3     Thermogravimetric analyses (TGA) were carried out in order to investigate the thermal behavior and composition of the magnetic nanocomposites. The loss of physisorbed water can be noticed for all materials on heating up from 25 to 140 °C (Figure 8). Two superimposed mass loss events can be noticed in the temperature range 140-650 °C for all the samples. These two thermal events roughly centered at 300 and 500 °C correspond to the combustion of organic groups and the oxidation of magnetite to Fe2O3. The oxidation of Fe3O4 to Fe2O3 after the heat treatment was confirmed visually, as the materials changed color from brown-black before the TG analysis to the characteristic red color of Fe2O3. The mass decrease up to 650 °C corresponds to 4.1, 6.6, and 8.5 % wt. for the Fe3O4-SBA15, Fe3O4-SBA15-COOH-0.05, and Fe3O4-SBA15-COOH-0.1 samples, respectively. All three samples exhibit a gradual, 1.1% mass loss above 650 °C, which can be ascribed to the dehydration of silanol groups [29]. The percent of organic material was calculated by sub- Thermogravimetric analyses (TGA) were carried out in order to investigate the thermal behavior and composition of the magnetic nanocomposites. The loss of physisorbed water can be noticed for all materials on heating up from 25 to 140 °C (Figure 8). Two superimposed mass loss events can be noticed in the temperature range 140-650 °C for all the samples. These two thermal events roughly centered at 300 and 500 °C correspond to the combustion of organic groups and the oxidation of magnetite to Fe2O3. The oxidation of Fe3O4 to Fe2O3 after the heat treatment was confirmed visually, as the materials changed color from brown-black before the TG analysis to the characteristic red color of Fe2O3. The mass decrease up to 650 °C corresponds to 4.1, 6.6, and 8.5 % wt. for the Fe3O4-SBA15, Fe3O4-SBA15-COOH-0.05, and Fe3O4-SBA15-COOH-0.1 samples, respectively. All three samples exhibit a gradual, 1.1% mass loss above 650 °C, which can be ascribed to the dehydration of silanol groups [29]. The percent of organic material was calculated by sub-   a superparamagnetic behavior of all samples. The calculated saturation magnetization values are similar: 13.3 emu/g for Fe3O4-SBA15, 13.0 emu/g for Fe3O4-SBA15-COOH-0.05, and 12.5 emu/g for Fe3O4-SBA15-COOH-0.1. The Ms values are smaller than that of bulk Fe3O4 (88-94 emu/g) due to the presence of the diamagnetic silica particles. However, these values indicate that the investigated nanocomposites can be isolated quickly and efficiently from solution under the influence of an external magnetic field. For the adsorption processes of pollutants from aqueous solutions, this property is particularly important.

Effect of pH on Dye Adsorption
The solution pH is a crucial factor for the adsorption process because it affects both the surface of the adsorbent and the adsorbate structural changes. The effect of pH on dye adsorption was studied in the range 2-11. According to the obtained results ( Figure 10), in both cases the adsorption capacities increase by increasing the pH and reach a maximum at pH 10.6. Consequently, this value was selected for the following investigations. Low removal efficiency in acidic solutions can be explained by the fact that at low pH values, the carboxyl groups are protonated; hence, the interaction with the cationic dyes' molecules is lower. As the pH increases, more and more carboxyl functional groups dissociate and the surface of the adsorbent becomes more negative; therefore, the electrostatic attraction between the adsorbent and dye molecules increases.

Effect of pH on Dye Adsorption
The solution pH is a crucial factor for the adsorption process because it affects both the surface of the adsorbent and the adsorbate structural changes. The effect of pH on dye adsorption was studied in the range 2-11. According to the obtained results ( Figure 10), in both cases the adsorption capacities increase by increasing the pH and reach a maximum at pH 10.6. Consequently, this value was selected for the following investigations. Low removal efficiency in acidic solutions can be explained by the fact that at low pH values, the carboxyl groups are protonated; hence, the interaction with the cationic dyes' molecules is lower. As the pH increases, more and more carboxyl functional groups dissociate and the surface of the adsorbent becomes more negative; therefore, the electrostatic attraction between the adsorbent and dye molecules increases.

Effect of Contact Time on Dye Adsorption
The adsorption of MB onto all three adsorbents is a rapid process, with the adsorption rate increasing quickly during the first 30 min, then slowing down gradually with time until reaching the equilibrium (after 240 min) ( Figure 11). This behavior can be explained as follows: during the early stage of adsorption, the number of active sites available on the external and internal surface of the adsorbent and the adsorbate concentration gradient is high, leading to a fast adsorption rate. In the second stage, when the most accessible surface sites tend to saturate, MB gradually diffuses to the active sites located on the internal surface of the mesoporous silica, in the smallest pores; hence, the adsorp-

Effect of Contact Time on Dye Adsorption
The adsorption of MB onto all three adsorbents is a rapid process, with the adsorption rate increasing quickly during the first 30 min, then slowing down gradually with time until reaching the equilibrium (after 240 min) ( Figure 11). This behavior can be explained as follows: during the early stage of adsorption, the number of active sites available on the external and internal surface of the adsorbent and the adsorbate concentration gradient is high, leading to a fast adsorption rate. In the second stage, when the most accessible surface sites tend to saturate, MB gradually diffuses to the active sites located on the internal surface of the mesoporous silica, in the smallest pores; hence, the adsorption rate decreases. After 240 min, the concentration of the MB solution remains almost unchanged, indicating that all the adsorption sites were saturated. In the case of MG, Fe 3 O 4 -SBA15-COOH-0.1 behaves similar as for MB, while for Fe 3 O 4 -SBA15 and Fe 3 O 4 -SBA15-COOH-0.05, the adsorption rate increases monotonically from the beginning of the process until the equilibrium (after 300 min). solutions.

Effect of Contact Time on Dye Adsorption
The adsorption of MB onto all three adsorbents is a rapid process, with the adsorption rate increasing quickly during the first 30 min, then slowing down gradually with time until reaching the equilibrium (after 240 min) ( Figure 11). This behavior can be explained as follows: during the early stage of adsorption, the number of active sites available on the external and internal surface of the adsorbent and the adsorbate concentration gradient is high, leading to a fast adsorption rate. In the second stage, when the most accessible surface sites tend to saturate, MB gradually diffuses to the active sites located on the internal surface of the mesoporous silica, in the smallest pores; hence, the adsorption rate decreases. After 240 min, the concentration of the MB solution remains almost unchanged, indicating that all the adsorption sites were saturated. In the case of MG, Fe3O4-SBA15-COOH-0.1 behaves similar as for MB, while for Fe3O4-SBA15 and Fe3O4-SBA15-COOH-0.05, the adsorption rate increases monotonically from the beginning of the process until the equilibrium (after 300 min).

Adsorption Kinetics
The kinetics of the adsorption process is essential to understand the adsorption mechanism and for optimization of the operating conditions in full-scale batch processes. The nonlinear forms of all these kinetic models are described by the following equations: -Pseudo-first-order model: -Pseudo-second-order model: -Intraparticle diffusion model: where Q e and Q t are the amount of dye adsorbed at equilibrium and at time t (mg g −1 ), k 1 is the rate constant of pseudo-first-order kinetics (min −1 ), k 2 is the rate constant of pseudo-second-order kinetics (g mg −1 min −1 ), k id is the intraparticle diffusion rate constant (g mg −1 min −1 ) and the intercept of the plot, and C reflects the boundary layer effect. Figures 12-14 show the PFO and PSO kinetic models fitting the data for MB and MG adsorption onto magnetic composites in single and binary solutions, while Tables 2 and 3 display the values of the calculated kinetic parameters using the PFO and PSO nonlinear models. As can be seen, the values of adjusted R 2 are higher for the PSO model than for the PFO model in all the cases, indicating that the PSO model is more suitable to describe the adsorption process of MB and MG, respectively, onto the magnetic nanocomposites.
In addition, the theoretical values of Q e (Q e cal ) correlate better with the experimental ones (Q e exp ), confirming that the adsorption process follows a PSO kinetic model and chemisorption is the rate-limiting step [30].
pseudo-second-order kinetics (g mg min ), kid is the intraparticle diffusion rate constant (g mg −1 min −1 ) and the intercept of the plot, and C reflects the boundary layer effect. Figures 12-14 show the PFO and PSO kinetic models fitting the data for MB and MG adsorption onto magnetic composites in single and binary solutions, while Tables 2 and 3 display the values of the calculated kinetic parameters using the PFO and PSO nonlinear models. As can be seen, the values of adjusted R 2 are higher for the PSO model than for the PFO model in all the cases, indicating that the PSO model is more suitable to describe the adsorption process of MB and MG, respectively, onto the magnetic nanocomposites. In addition, the theoretical values of Qe (Qe cal) correlate better with the experimental ones (Qe exp), confirming that the adsorption process follows a PSO kinetic model and chemisorption is the rate-limiting step [30]. The phenomena that limit the sorption mechanism are usually determined using the Weber and Morris kinetic intraparticle diffusion model. According to this model, if a straight line passing through the origin is generated from the plot of Equation (4), it can be said that the adsorption mechanism involves intraparticle diffusion of the species [31]. The slope of the linear curve is the rate constant of the intraparticle diffusion process. In our study, when the kinetic data obtained for MB adsorption onto the magnetic composites were analyzed using the Weber and Morris intraparticle diffusion model, it was observed that the plot did not pass through the origin, indicating that intraparticle diffusion was not the only rate-limiting step. According to this model, the adsorption of MB occurs in three stages revealed through a multilinear curve composed of three segments (Figure 15). In this case, a piecewise linear regression was applied to the experimental data using a Microsoft Excel worksheet developed by Malash and El-Khaiary [32]. The results are presented in Figure 15 and Table 4. In the first stage, the adsorption is limited to the external diffusion of the adsorbate. The second stage consists of gradual adsorption of the adsorbate being limited by intraparticle diffusion, while the third one corresponds to an equilibrium phase during which the adsorption capacity remains stable [33]. This pattern shows that both external mass transfer and intraparticle diffusion are involved in adsorption.   [32]. The results are presented in Figure 15 and Table 4. In the first stage, the adsorption is limited to the external diffusion of the adsorbate. The second stage consists of gradual adsorption of the adsorbate being limited by intraparticle diffusion, while the third one corresponds to an equilibrium phase during which the adsorption capacity remains stable [33]. This pattern shows that both external mass transfer and intraparticle diffusion are involved in adsorption.
(a) (b) (c)  In the case of MG adsorption, the variation in Q t versus t 0.5 is shown in Figure 16. The slope of the straight line (k id ), the intercept (C), and the regression coefficients are given in Table 4. It can be observed that the straight line did not pass through the origin which means that intraparticle diffusion is not the only rate-limiting step and the intercept has negative values. Previous studies that reported a negative value of C suggested that external film diffusion, in addition to intraparticle diffusion, limited the adsorption rate [34]. Therefore, in our case, the negative values of C constant might be interpreted as an external film diffusion resistance that led to the time lag for the MG adsorption [35]. In the case of MG adsorption, the variation in Qt versus t 0.5 is shown in Figure 16. The slope of the straight line (kid), the intercept (C), and the regression coefficients are given in Table 4. It can be observed that the straight line did not pass through the origin which means that intraparticle diffusion is not the only rate-limiting step and the intercept has negative values. Previous studies that reported a negative value of C suggested that external film diffusion, in addition to intraparticle diffusion, limited the adsorption rate [34]. Therefore, in our case, the negative values of C constant might be interpreted as an external film diffusion resistance that led to the time lag for the MG adsorption [35].

Adsorption Isotherms
The Langmuir (Equation (5)) and Freundlich (Equation (6)) isotherm models are the most known isotherm models that can describe the adsorption equilibrium between the concentration of the dye in bulk solution and the quantity of the dye adsorbed on the adsorbent surface at a given temperature. The Langmuir model assumes uniform adsorption energies distributed on the adsorbent surface, whereas the Freundlich isotherm model assumes multilayer adsorption on a heterogeneous surface with nonuniform affinities and adsorption energies [36]. In our study, the equilibrium data were modelled using both equilibrium isotherm models whose nonlinear equations are the following:

Adsorption Isotherms
The Langmuir (Equation (5)) and Freundlich (Equation (6)) isotherm models are the most known isotherm models that can describe the adsorption equilibrium between the concentration of the dye in bulk solution and the quantity of the dye adsorbed on the adsorbent surface at a given temperature. The Langmuir model assumes uniform adsorption energies distributed on the adsorbent surface, whereas the Freundlich isotherm model assumes multilayer adsorption on a heterogeneous surface with nonuniform affinities and adsorption energies [36]. In our study, the equilibrium data were modelled using both equilibrium isotherm models whose nonlinear equations are the following: where C e is the equilibrium concentration of the solute in solution (mg L −1 ), K L is the equilibrium constant of the Langmuir model related to the adsorption energy (L mg −1 ), Q e is the adsorption capacity at equilibrium (mg g −1 ), Q max is the maximum adsorption capacity (mg g −1 ), and K F and 1/n are Freundlich isotherm parameters (adsorption capacity (mg g −1 ) and intensity). Figures S3 and S4 and Table 5 illustrate the adsorption isotherms and the calculated parameters. The correlation coefficient (R 2 ) and the Akaike's information criterion (AIC) were used to confirm the goodness of fit. Lower AIC values imply that the respective model is more likely than the alternative model to characterize the sorption process [37]. Based on these parameters, it is clear that the Langmuir model fits better the experimental data for MB adsorption, whereas the Freundlich model describes better the adsorption process of MG at equilibrium. The favorability of the adsorption process was estimated from the values of the separation factor (R L ), using the following equation: where C 0 (mg L −1 ) is the initial dye concentration, and K L is the Langmuir equilibrium constant. When R L is between 0 and 1 the adsorption process is considered as favorable, for R L > 1 the adsorption is unfavorable, while for R L = 1 the process is linear [38]. In this study, the calculated values of R L fall between 0 and 1, which shows the favorability of the adsorption process.
The modified Langmuir isotherm model can be used to determine the competitive adsorption capacity of dyes in binary solutions [39]. Equation (8) describes this model mathematically.
Q e,D1 = Q max,D1 K L,D1 C e,D1 1 + K L,D1 C e,D1 + K L,D2 C e,D2 The linearized form of Equation (8) is the following: For dye 2 (D2), the linearized equation is: where C e,D1 , C e,D2 , Q e,D1 , and Q e,D2 are the equilibrium concentration and the equilibrium adsorption capacity of dye 1 (D1) and dye 2 (D2) in binary solutions; K L,D1 and K L,D2 are the Langmuir constants characteristics for the dye adsorption from single solutions; and Q max,D1 and Q max,D2 are the maximum adsorption capacities of the magnetic composite for D1 and D2 in binary solution. Q max,D1 and Q max,D2 can be obtained by graphing Equations (9) and (10), respectively.
The ratio Q max,binary Q max,single offers information about the dynamics of dye adsorption in binary solutions [40]. The two adsorbates have a synergistic effect when this ratio is supraunitary, with the mixture's effect stronger than the individual adsorbates' effect. When the ratio is less than 1, the two adsorbates have an antagonistic effect, with the mixture having a weaker effect than the individual adsorbates. When the ratio is equal to 1, the combination has no influence on the dyes' adsorption [40]. In Table 6, the values of Q max,MB and Q max,MG , as well as the ratio Q max,binary /Q max,single for each magnetic nanocomposite investigated in this study, are displayed. Only in the case of MB adsorption onto Fe 3 O 4 -SBA15 is the Q max,binary /Q max,single ratio equal to 1, implying that the mixture has no influence on the adsorption of each adsorbate. In all other cases, the ratio is subunitary, which means that the adsorption of each dye is hindered by the presence of the other one.

Adsorption Mechanism
In general, a high adsorption capacity is linked to a large surface area and pore volume of the adsorbent. For the magnetic nanocomposites investigated in this study, these two parameters have close values (Table 1); therefore, the differences between their adsorption capacities depend on other factors, such as the size of the adsorbate molecule and existing functional groups that may cause weaker or stronger interactions with the adsorbent surface. The structure and surface chemistry of the adsorbent also play a major role in dye molecule adsorption. During the adsorption process, a variety of interactions could occur such as hydrogen bonding, electrostatic interactions, van der Waals forces, and π-π interactions [41]. The driving force of adsorption, according to some researchers, is the molecular geometry and surface charge [42]. According to zeta potential measurements, the magnetic nanocomposites investigated in our study have negative charges (−32.  (Table 1). It can be noted that zeta potential values increase as the content of the COOH groups increases. These negative charges indicate that these nanocomposites present more favorable adsorption for cationic dyes and a very good dispersibility in water, which is very important for the adsorption of pollutants from aqueous solutions. The electrostatic interaction seems to be the dominant mechanism of adsorption, but the hydrogen bonding could also be involved in the adsorption process. A schematic representation of these interactions is shown in Figure 17. The analysis of equilibrium data showed that MB dye is better adsorbed by the studied magnetic nanocomposites than MG dye. These experimental results could be explained by the molecular geometry of the dye molecule. The MB molecule, having a planar structure, favors the adsorption via a face-to-face conformation. On the other hand, nonplanar molecules such as MG are kept aside from the adsorbent surface due to the spatial constraint, resulting in low interactions with the adsorbent surface. This behavior was also observed by other researchers who investigated dye adsorption by modified multiwalled carbon nanotubes. They found that molecules with a greater charge and planar structure have a higher adsorption affinity [42]. Table 7 shows a comparison of the performance of the magnetic nanocomposites investigated in this study with other similar adsorbents reported in the literature. As can be seen, their adsorption capacity is comparable to or even superior to that of the other adsorbents. Therefore, these novel magnetic nanocomposites might be considered as effective adsorbents to remove cationic dyes and other contaminants from wastewater.

Desorption Studies
The desorption and regeneration capacity of the adsorbents are of crucial importance when assessing their industrial applications. In our study ethanol 94% has been tested as desorbing agent. As can be seen in Figure 18, after seven repeated adsorption-desorption cycles, the adsorption capacity of the adsorbents exhibited insignificant decay. Therefore, the nanocomposites exhibited superior recycling stability for the removal of organic dyes MB and MG from aqueous solutions.  The desorption and regeneration capacity of the adsorbents are of crucial importance when assessing their industrial applications. In our study ethanol 94% has been tested as desorbing agent. As can be seen in Figure 18, after seven repeated adsorption-desorption cycles, the adsorption capacity of the adsorbents exhibited insignificant decay. Therefore, the nanocomposites exhibited superior recycling stability for the removal of organic dyes MB and MG from aqueous solutions.

Conclusions
In this study, three novel magnetic nanocomposites based on carboxyl-functionalized SBA-15 silica and magnetite nanoparticles, with relatively high surface area and total

Conclusions
In this study, three novel magnetic nanocomposites based on carboxyl-functionalized SBA-15 silica and magnetite nanoparticles, with relatively high surface area and total pore volume, were prepared through a simple method, and their adsorption capacity for methylene blue and malachite green G from single and binary aqueous solutions was investigated comparatively. An increase in the sorption capacity for both dyes by increasing the amount of carboxyl groups on the adsorbent surface was observed. The results showed that a pH value of 10.6 is the most favorable for dye adsorption. The kinetic studies revealed that dye adsorption onto all three adsorbents followed a pseudo-second-order kinetics model, the electrostatic interaction being the dominant mechanism of adsorption. The hydrogen bonding seems to be also involved in the adsorption process. The results of the intraparticle diffusion model indicated that the adsorption process is significantly influenced by external mass transfer and intraparticle diffusion. The equilibrium adsorption data recorded for MB were best fitted by the Langmuir model, whereas for MG, the Freundlich model described better the adsorption process at equilibrium. The maximum adsorption capacities were determined to be 239.17/30.73 mg g −1 (MB/MG) for Fe 3 O 4 -SBA15, 254.58/39.28 mg g −1 (MB/MG) for Fe 3 O 4 -SBA15-COOH-0.05, and 256.09/ 126.55 mg g −1 (MB/MG) for Fe 3 O 4 -SBA15-COOH-0.1. These values are higher than most of those reported in the literature for similar materials. The obtained nanocomposites proved to be very good adsorbents for cationic dyes in single and binary solutions. Their high recycling stability, regeneration capacity, and efficient reuse in multiple cycles, recommends these nanocomposites as promising materials for wastewater treatment.