Magnetic solid phase adsorption, preconcentration and determination of methyl orange in water samples using silica coated magnetic nanoparticles and central composite design

Reagents and materials

All chemicals and reagents used in this work were of analytical grade. Iron nitrate Fe(NO ₃ ) ₃ ·9H ₂ O (99 %), tetra ethoxy silane (TEOS) (98 %) and oxalic acid H ₂ C ₂ O ₄ ·2H ₂ O were purchased from Merck (Darmstadt, Germany). MO was prepared from Merck (Darmstadt, Germany). Structural formula of MO has been shown in Scheme  ¹ . Double distilled water was used throughout the study. The 200.0 mg L ⁻¹ stock solution of MO was prepared in double distilled water and experimental solutions of the desired concentrations were obtained by successive dilutions of the stock solution with double distilled water. The initial pH was adjusted with 0.1 mol L ⁻¹ solutions of HCl or NaOH. All the adsorption experiments were carried out at room temperature.

Scheme 1

Instrumentation

Recording spectra was performed by an Agilent model 8453 spectrophotometer with diode array detector. A Jenway 3345 ion-meter was used for pH measurements.

A Philips X′ Pert (40 kV, 30 mA) X-ray diffractometer, using a Cu Kα radiation source ( λ  = 1.542 Å) and a nickel filter in the 2 θ range of 4°–70° was used to record X-ray diffractometer (XRD) pattern of the calcined sample.

The specific surface area (using BET and BJH methods), the total pore volume and the mean pore diameter were measured using a N ₂ adsorption–desorption isotherm at liquid nitrogen temperature (−196 °C) using a NOVA 2200 instrument (Quantachrome, USA). Prior to the adsorption–desorption measurements, all the samples were degassed at 110 °C in a N ₂ flow for 3 h to remove the moisture and other adsorbates.

The morphologies of the prepared nanoparticles and their precursors were observed by means of an EM-3200 scanning electron microscope (KYKY CO., China).

For recording FTIR spectrum, an Alpha FTIR spectrometer (Bruker, Germany) was employed.

Dye adsorption experiments

Batch-mode adsorption studies were carried out by adding 10.0 mg adsorbent and 10.0 mL dye solution of known concentration (1.0 mg L ⁻¹ ) to a beaker. Then, pH of the solutions was adjusted to the desired value. The mixture solutions were shacked for appropriate adsorption time at 25 °C. After dye adsorption, SMNPs were quickly separated from the sample solution using a magnet. The following equation was used to calculate the dye adsorption efficiency in the treatment experiments: where Ci and Cr are the initial and residual concentrations of the dye in the solution, respectively.

Synthesis of SiO2-coated Fe3O4 magnetic nanoparticles

The SMNPs were prepared using sol–gel method [ ³¹ – ³³ ]. For preparation, 5.252 g of Fe(NO ₃ ) ₃ ·9H ₂ O, 2.395 g of H ₂ C ₂ O ₄ ·2H ₂ O and 4.5 mL of TEOS were separately dissolved in ethanol. These solutions were heated up to 50 °C and stirred for 20 min. TEOS was added to the iron nitrate followed by addition of oxalic acid under strong stirring at 60 °C for 2 h. The precipitate composed of iron oxalate and TEOS is progressively hydrolyzed by hydration water of iron nitrate and mainly oxalic acid, according to the following scheme:

In acidic conditions (pH ≈ 1), Si(OH) ₄ is condensed with other materials to a homogeneous gel. The monolithic gel was dried at 110 °C in vacuum for 16 h. Finally, the dried powder was calcined at 450 °C for 6 h to produce solid magnetic nanoparticle.

Characterization of the SiO2-coated Fe3O4 magnetic nanoparticles

We began characterization studies of the synthesized MNPs with XRD technique. XRD pattern of the synthesized MNPs have been shown in Fig. ¹ . The background of XRD pattern was subtracted from 2 θ  = of 4°–70° with smoothing of 50 %. Smoothing of 50 % means that the intensities of the all peaks have been decreased by 50 % in comparison with the original peaks. The peak positions and width were determined with Match 1.11b software. We assigned XRD pattern according to the data of JCPDS for detection of the actual phases and we added the label for all phases in the XRD pattern. The phases identified based on the XRD studies for this sample were cubic Fe ₂ SiO ₄ . Widths of all major peaks in XRD pattern were determined and the average was used for calculation of the size of the crystallite of the synthesized sample by Debye–Scherrer equation [ ³⁴ ]: where β is the breadth of the observed diffraction line at its half intensity maximum, K is the so-called shape factor which usually takes a value of about 0.9 and λ is the wavelength of the X-ray source used in the XRD. Using the above equation, the mean crystallite size ( D _c ) of the synthesized MNPs was calculated to be 48 nm.

Fig. 1

The FTIR spectrum of the synthesized MNPs was recorded and shown in Fig.  ² . The spectrum shows the core shell of the particles with Si–O on the surface. The Broad peak at 3,447 cm ⁻¹ could be correlated to the vibration of structural OH [ ³⁵ ]. The bands at 1,083, 800 and 452 cm ⁻¹ are attributed to the asymmetric Si–O–Si stretching vibration, the symmetric Si–O–Si stretching vibration and the O–Si–O shearing vibration, respectively [ ³⁶ , ³⁷ ]. The bands around 1,000 cm ⁻¹ were assigned to the asymmetric Fe–O–Si stretching vibration [ ³⁶ , ³⁸ , ³⁹ ]. These results suggested that the Fe–SiO ₂ interaction exists in the catalyst in the form of Fe–O–Si structure. The band at 551 cm ⁻¹ can be assigned to the Fe–O stretching in Fe–O–Si bonds [ ³⁵ ].

Fig. 2

To reveal the fine details of the structure of the synthesized MNPs, SEM study of both precursor and calcined SMNPs was performed. The SEM images recorded are shown in Fig.  ³ . The image obtained from the precursor depicts several larger agglomerations of particles (Fig.  ³ a) and shows that this material has a less dense and homogeneous morphology. After calcination at 450 °C for 6 h and heating rate of 3 °C min ⁻¹ , the morphological features became different from the precursor sample and the agglomerate size reduced greatly (Fig.  ³ b). It may be attributed to the covering of calcined magnetic nanoparticle surface by small crystallite of SiO ₂ , in agreement with XRD results.

Fig. 3

Central composite design for optimization of the factors

In experimental design, the level of all factors is changed from one experiment to the next, simultaneously. The reason for this is that factors can interact with each other and the ideal value for one of them can depend on the values of the others. In this work, we used central composite design (CCD). CCD as an experimental design method is commonly used for process analysis and modeling [ ⁴⁰ , ⁴¹ ]. It is assumed that the central point for each factor is 0, and the design is symmetric around this [ ⁴² ]. CCD including the factors, their levels and the result of each experiment is shown in Table  ¹ . Before performing experimental design, preliminary tests showed that in longer times, the adsorption of the dye is higher. Therefore, 20 min was used as the lower limit of the time for designing experiments. Concentration of dye in these experiments was 1.0 mg L ⁻¹ .

The experiments in Table  ¹ were performed and their experimental responses were obtained. Here, the percent of the adsorption of MO on the adsorbent was used as experimental response. Analysis of variance for the results has been collected in Table  ² . It can be inferred from the results of ANOVA in Table  ² that pH is significant factor in the adsorption of MO at 95 % confidence level (its p value is smaller than α  = 0.05 for 95 % confidence level and t _statistics = 4.38). Parameter “ p ” is the probability (from 1) factor that affects the response by chance. Therefore, the lower is p , the more significant is the factor. This is compared with a critical value such as 0.05 (for 95 % confidence level). The statistical parameter “ t_ statistics” for a factor is calculated by dividing its estimated coefficient to its standard error. Therefore, a larger value of t_ statistics shows the significance of the corresponding factor. Among the parameters studied, mgMNP has the least influence in the adsorption of MO on the MNPs ( p  ≫ 0.05). Time ( t ) is not also an important factor. This was experimentally observed. Among the squared and interaction terms, pH × pH is statistically important based on the p values. The F value of the regression is relatively high ( F  = 5.11, p  = 0.001). This indicates that the regression is significant.

Three-dimensional graphs will give us a clear insight about the effect of each factor. For this purpose, the response was plotted based on the polynomial function with the coefficients in Table  ² versus the factors based on the ranges in Table  ¹ . The resulted three-dimensional response surfaces have been shown in Fig.  ⁴ . These figures show the relationship between two factors and response (adsorption %) at center level of the third factor. From the Fig.  ⁴ , it is clear that at intermediate pHs and times and with larger amounts of MNPs, the adsorption is higher. Moreover, curvatures in the plots in Fig.  ⁴ a, c show the probable importance of the second-order effect of pH. ANOVA table (Table  ² ) confirms this since the p value for the term pH × pH is 0.001.

Fig. 4

In the next step, response surface optimization was used to explore the optimum values of the factors. Response optimization showed that the adsorption % will be maximum at t  = 120 min, mgMNP = 10.0 mg and pH 2.66.

Mechanism of the interaction

Results of experimental design showed that in acidic medium (pH 2.66) adsorption is higher. Si–OH groups formed by interaction of SiO ₂ with water are capable of proton transfer. This results in the variation in the surface charge of the MNPs with pH. At lower pHs, surface sites are protonated and the surface becomes positively charged. However, at higher pHs, the surface of the MNPs is negatively charged due to the partial ionization of Si–OH.

At low pH

At high pH

MO is an ionic dye with –SO ₃ Na group. At low pH, the negative charge of the sulfonic group predominates and it is responsible for the electrostatic interaction with the adsorbent. At more acidic pHs, SO ₃ ⁻ groups can be protonated and the surface becomes positively charged. Therefore, it is expected that the adsorption of MO on the MNPs decreases. At higher pHs, both MO and the surface of the MNPs are predominately negatively charged. This will result in the decrease in adsorption in higher pHs. These explanations are clearly confirmed by Fig.  ⁴ c.

Study of the kinetics of adsorption

Study of the kinetics of dye adsorption onto SMNPs is required for selection of the optimum operating conditions for the full-scale batch processes. The kinetic parameters which are helpful for the prediction of the adsorption rate give important information for designing and modeling of the adsorption processes [ ⁴³ ].

Kinetic studies were performed in a 15 mL glass beaker where 10.0 mg of SMNPs was added to 10.0 mL of the dye solution with concentrations between 10.0 and 30.0 mg L ⁻¹ at room temperature and in pH 2.66. Then, using the shaker the mixture was shaken at 250 rpm to ensure equilibration between solution and adsorbent. At time t  = 0 and equilibrium, dye concentration was measured by UV–Vis spectrophotometry at 505 nm. The amount of adsorption at equilibrium, qt (mg g ⁻¹ ), was calculated by: where Ci and Ct (in mg L ⁻¹ ) are concentrations of dye in solution at t  = 0 and equilibrium after time t of incubation, respectively. V is the volume of the solution (in L) and W is the mass of SMNPs (in g).

Pseudo-second-order model

The rate of removal of MO was very fast during the initial stages of the adsorption process. The kinetic data for adsorption of the dye onto SMNPs were firstly analyzed using pseudo-second-order model. The results showed that the kinetics of adsorption obeys pseudo-second-order model. The kinetics of the adsorption of dye with different initial concentrations was treated by Ho’s pseudo-second-order rate equation [ ⁴⁴ ]: where q _t and q _e are the amount of adsorbed dye at time t and at equilibrium, respectively. k ₂ is the pseudo-second-order rate constant. Plots of the kinetics data based on the pseudo-second-order kinetic model are shown in Fig.  ⁵ . The constants of the pseudo-second-order rate equation for different initial concentrations are shown in Table  ³ .

Fig. 5

Table 3

Initial concentration (mg L−1)	Equation	qe,exp (mg g−1)	qe,calc (mg g−1)	k (g mg−1 min−1)	R2
10	t/qt = 0.1222t + 0.2091	7.96	8.18	0.07	0.9999
20	t/qt = 0.0703t + 0.1213	13.88	14.22	0.04	0.9999
30	t/qt = 0.0570t + 0.0799	17.09	17.54	0.04	0.9999

Equilibrium isotherm models

Equilibrium isotherm models were used to describe the experimental adsorption data. The parameters obtained from different models provide important information about the adsorption mechanism, the surface properties and affinity of the adsorbent [ ⁴⁵ ]. The common models for investigation of the adsorption isotherm (Langmuir and Freundlich) were also fitted to the experimental data [ ²³ ].

Langmuir’s model does not take into account the variation in the adsorption energy, but it is the simplest description of the adsorption process. It is based on the physical hypotheses that the maximum adsorption capacity consists of a monolayer adsorption, there are no interactions between adsorbed molecules and adsorption energy is distributed homogeneously over the entire coverage surface [ ⁴³ ]. The equilibrium adsorption isotherm was determined using batch studies with different initial concentrations of MO (10–100 mg L ⁻¹ ) at 25 °C and in pH 2.66.

The linearized form of the Langmuir isotherm is expressed as [ ⁴⁶ ]: where q _max (in mg g ⁻¹ ) is the maximum amount of the adsorbed dye which corresponds to the complete monolayer coverage and illustrates the maximum value of q _e that can be attained as Ce increases. K _L (in L mg ⁻¹ ) is the Langmuir adsorption equilibrium constant which is related to the energy of adsorption. Values of q _max and b ( K _L / q _max ) are determined by the linear regression plot of (Ce/ q _e ) versus Ce. Adsorption data plotted based on the Langmuir isotherm are shown in Fig.  ⁶ .

Fig. 6

The Freundlich isotherm model is an empirical equation that describes the surface heterogeneity of the adsorbent. It considers multilayer adsorption with a heterogeneous energetic distribution of active sites accompanied by interactions between adsorbed molecules [ ⁴⁷ ]. The linear form of the Freundlich isotherm is: where C _e is the equilibrium concentration (in mg L ⁻¹ ), q _e is the amount of the adsorbed dye in equilibrium (mg g ⁻¹ ) and finally, K _f [in (mg g ⁻¹ ) (mg L ⁻¹ ) ⁿ ] and 1/ n are Freundlich constants which depend on the temperature and the given adsorbent–adsorbate couple. Parameter n is related to the adsorption energy distribution and K _f indicates the adsorption capacity. The values of K _f and 1/ n can be calculated by the plotting Ln ( q _e ) versus Ln ( C _e ). The intercept of the resulted line is Ln ( K _f ) and its slope is 1/ n . Value of 1/ n indicates that the adsorption intensity of dye onto the adsorbent (or surface heterogeneity) becomes more heterogeneous as its value gets closer to zero. A value for 1/ n below 1 indicates a normal Freundlich isotherm while 1/ n above 1 is indicative of the cooperative adsorption [ ²⁴ ]. Adsorption data plotted based on the Freundlich isotherm are shown in Fig.  ⁷ .

Fig. 7

It is generally accepted that under constant temperature, n increases with decreasing adsorption energy. This implies that the larger the n value, the stronger the adsorption intensity [ ⁴⁸ ]. Values of n  > 1 represent favorable adsorption conditions.

The essential feature of the Langmuir isotherm can be expressed in terms of a dimensionless constant separation factor ( R _L ) given by the following equation [ ⁴³ ]: where a _L parameter is a coefficient related to the energy of the adsorption which increases by increasing the strength of the adsorption bond. The adsorption process can be defined as irreversible ( R _L  = 0), favorable (0 <  R _L  < 1), linear ( R _L  = 1) or unfavorable ( R _L  > 1) in terms of R _L [ ²⁰ ] _.

The calculated parameters of Langmuir and Freundlich isotherms and the corresponding correlation coefficients ( R ) are listed in Table  ⁴ . These results show that experimental data follow the Freundlich model. Since the Freundlich adsorption model is based on the assumptions that all surface sites are different, the adsorption of MO occurs through different adsorption sites. However, the data fit the Langmuir isotherm, satisfactorily.

Table 4

aL (L mg−1)	KL (L g−1)	qmax| = KL/aL| (mg g−1)	RL	r
Langmuir isotherm
0.00082	0.04356	53.19	0.924	0.9219

Kf	n	r
Freundlich isotherm
5.087	1.997	0.9896

The calculated value of R _L for adsorption of 100.0 mg L ⁻¹ solution of MO is 0.924. Thus, the adsorption of dye onto SMNPs is favorable. The calculated n parameter is about 2 (Table  ⁴ ) which implies that adsorption is favorable. In most cases, the exponent between 1 <  n  < 10 shows beneficial adsorption [ ⁴³ ].

In a work published by Deligeer et al. [ ¹² ], γ-Fe ₂ O ₃ /SiO ₂ nanocomposite was synthesized and used to adsorb MO. Adsorption of MO onto this nanocomposite obeyed Freundlich isotherm, too [ ¹² ]. Calculated q _m based on the Langmuir isotherm in that work was reported as 476 mg g ⁻¹ which is higher than q _m obtained here. However, the range of the concentrations examined for adsorption is much lower in this work. Other adsorbents with active sites which are mostly SiO ₂ can be found. For example, q _m values equivalent to 333.33 [ ¹⁷ ], 24 [ ⁵ ], 33.8 [ ⁸ ] and 66.09 mg g ⁻¹ [ ¹⁰ ] have been reported. Therefore, it is seen that the q _m value obtained here is comparable with most of the reported ones.

Analytical studies

Desorption and regeneration

Reversibility of the adsorption of MO onto the SMNPs allows the regeneration or activation of the SMNPs for new adsorption experiments. Desorption of the dye from the SMNPs was studied using different solvents. Desorption process was separately examined by adding 2.0 mL volume of ethanol, pure acetic acid, and HCl and NaOH solutions with concentrations of 0.1 mol L ⁻¹ –0.01 g of the adsorbent loaded with MO. SMNPs were collected magnetically from the solution. Concentration of dye in the desorbed solution was measured by spectrophotometric method. Figure  ⁸ shows the percentage of the recovered dye. It can be concluded from Fig.  ⁸ that a 2.0 mL volume of 0.1 mol L ⁻¹ NaOH solution is the most effective eluent for desorption of MO from SMNPs. The results showed that desorption efficiencies higher than 99 % can be achieved in a short time of 3 min and in a single-step elution using 2.0 mL of 0.1 mol L ⁻¹ NaOH. Therefore, the dye could be desorbed from the loaded SMNPs by changing the pH of the solution to alkaline range. The SMNPs were washed with deionized water and reused for two successive adsorption processes with adsorption efficiencies higher than 90 % (Fig.  ⁹ ). In higher removal cycles, removal efficiency decreases. This may be due to oxidation, losing and/or dissolving some amounts of the adsorbent during the successive steps.

Fig. 8

Fig. 9