Journal of Molecular Liquids 212 (2015) 675–685

Contents lists available at ScienceDirect

Journal of Molecular Liquids

j ourna l homepage: www.e lsev ie r .com/ locate /mol l iq

Removal of Safranin dye from aqueous solution using magneticmesoporous clay: Optimization study

M. Fayazi a,b,⁎, D. Afzali c, M.A. Taher d, A. Mostafavi d, V.K. Gupta e,f,g

a Mineral Industries Research Center, Shahid Bahonar University of Kerman, Kerman, Iranb Young Researchers Society, Shahid Bahonar University of Kerman, Kerman, Iranc Department of Environment, Institute of Science and High Technology and Environmental Sciences, Graduate University of Advanced Technology, Kerman, Irand Department of Chemistry, Faculty of Sciences, Shahid Bahonar University of Kerman, Kerman, Irane Department of Chemistry, Indian Institute of Technology Roorkee, Roorkee 247667, Indiaf Center for Environment and Water, Research Institute, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabiag Department of Applied Chemistry, University of Johannesburg, Johannesburg, South Africa

⁎ Corresponding author at: Mineral Industries ReseUniversity of Kerman, Kerman, Iran.

E-mail address: maryam.fayazi@yahoo.com (M. Fayaz

http://dx.doi.org/10.1016/j.molliq.2015.09.0450167-7322/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 17 June 2015Received in revised form 22 September 2015Accepted 26 September 2015Available online xxxx

Keywords:Magnetic adsorbentSafranin removalCentral composite designSepioliteResponse surface methodology

Iron oxide/sepiolite magnetite composite (MSep) was prepared by a chemical precipitation method. The com-posite was characterized by using X-ray diffraction, scanning electron microscopy, Fourier transform infraredspectroscopy and specific surface area analysis. The response surface methodology (RSM) based on central com-posite design (CCD)was successfully applied to the optimization of the Safranin removal process. Three indepen-dent variables namely initial pH, dye ion concentration and adsorbent dosage were investigated. Analysis ofvariance (ANOVA) of the quadratic model suggested that the predicted values were in good agreement with ex-perimental data. Detailed kinetic and equilibrium studies were performed for liquid phase adsorption of SafraninusingMSep. The adsorption process could bewell described by Langmuir isotherm and themaximummonolayeradsorption capacitywas calculated as 18.48mgg−1. The adsorption kinetics was evaluated by pseudo-first-orderand pseudo-second-ordermodels; pseudo-second-ordermodelwas found to describe the process better. The ad-sorptionwas analyzed thermodynamically and the results revealed that the adsorption processwas spontaneousand endothermic.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Dyes and pigments are extensively used in the textile, leather, plas-tics, printing, rubber, pharmaceutical, cosmetic, food, paper and carpetindustries to color their product. There are more than 10,000 differentdyes weighing approximately 7 × 105 tons that are produced annuallyfor various industrial processes [1]. The dyes generally have complex ar-omatic structure and thus most of them are highly resistant to break-down by chemical, physical, and biological treatments [2,3]. Thedischarge of dye-bearing wastewater into water bodies possesses a se-rious pollution problemas the dyes givewater undesirable color, reducesunlight penetration and gas solubility in water [4,5]. Safranin (3,7-di-methyl-10-phenylphenazin-10-ium-2,8-diamine chloride) is a water-soluble organic dye, widely used in textile industries [6]. However, Saf-ranin can cause eye burns which may be responsible for permanent in-jury to the cornea and conjunctiva in human and rabbit eyes [7]. Contactwith Safranin dye also causes skin and respiratory tract irritation [8]. Be-cause of this, industrial wastewaters containing such dye need to be

arch Center, Shahid Bahonar

i).

treated before being delivered to the environment [9,10]. A widerange of physical and chemical processes such as flocculation, adsorp-tion,membrane filtration, coagulation, precipitation, ozonation, electro-chemical techniques, and fungal decolonization have been investigatedextensively for removing dyes from aquatic bodies [11,12]. Amongthese, liquid phase adsorption has been found to be superior to othertechniques for removal of colors, odor, oils, and organic pollutantsfrom process or waste effluents. This is attributed to its initial lowcost, high efficiency, simplicity of design, and ease of operation [13].The major drawback of this method is the high price of adsorbentsthat increase the cost of treatment. A number of scientists for this pur-pose have used different adsorbents such as charcoal [14], zeolites[15], bagasse [16], fly ash [17], clay [18] and sawdust [19].

Sepiolite (magnesium hydro-silicate) is a natural clay mineral char-acterized by its fibrousmorphology and intracrystalline channelswith aunit cell formula of Si12Mg8O30(OH)4(OH2)4·8H2O [20]. The hollowneedlelike crystal structure of sepiolite is responsible for its uniquephysiochemical properties. The average size of fibrous sepiolite crystalsis 800× 25 × 4 nm,which results in a solidwith an external surface areaof the same order of magnitude as the area of macroporous [21]. Likeother clay minerals, natural sepiolite has an electronegative surface be-cause of isomorphic substitutions. In addition, the abundant molecular

676 M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

sized channels allow the penetration of organic and inorganic speciesinto the structure of sepiolite and assign sepiolite important industrialapplications [22–24].

During the past few years, magnetic nanoparticles (MNPs) as an ef-ficient adsorbent with large specific surface area and small diffusion re-sistance have been recognized [25,26]. The magnetic separationprovides a desirable path for online separation. A distinctive superiorityof this technique is that the MNPs with affinity to target species can bereadily isolated from sample solutions using an external magnetic fieldwithout additional filtration or centrifugation steps.

Response surface methodology (RSM) is a well known statistics-based procedure for designing experiments, understanding the effectsof different factors and their interactions on targeted response, buildingmodels and finding out optimum conditions [27–29]. The use of RSMhas been accentuated for developing the complex processes, optimizingtheir performance, and improving design of newproducts. Themain ad-vantage of RSM is the reduced number of experimental trials needed toevaluate multiple parameters and their interactions [30]. Central com-posite design (CCD) is the most frequently used method of RSM,which is suitable for fitting a quadratic surface and helps optimize theeffective parameters with a minimum experimental runs, in additionto analyzing the interaction between parameters [31].

The focus of this research was to explore the feasibility of new kindof Fe3O4/sepiolite magnetic nanocomposite (MSep) being utilized in re-moval of Safranin as a cationic dye from aqueous solutions and also op-timization of the process variables using the response surfacemodelingapproach. CCD was selected to study the individual and synergetic ef-fects of factors such as Safranin concentration (mg L−1), adsorbent dos-age (g) and pH on the percentage removal of dye as response.

2. Experimental

2.1. Instruments and reagents

All reagents and chemicals used in the study were of analytical re-agent grade. Safranin dye (C.I.: 50240; chemical formula: C20H19ClN4;M.W.: 350.84 g mol−1, λmax: 520 nm) was procured from Merck. Thechemical structure of the dye is shown in Fig. 1. Safranin stock solutionwith an initial concentration of 1000mgL−1was preparedbydissolvingthe required amount in double-distilled water. The test solutions wereprepared by diluting the stock solution to the desired concentrations.The pH was adjusted and measured using a 713 pH-mV meter(Metrohm, Switzerland) at the laboratory ambient temperature. Absor-bance spectra of Safranin were acquired with a Cary 50 single detectordouble beam in-time spectrophotometer (Varian, Australia). A newlyfound Iranian sepiolite (Sep) sample with a particle size of ≤0.075 mm(200-mesh) from Yazd region (central Iran) was applied in this study[32]. The mineral sample was powdered and sieved using a 0.05 mm

Fig. 1. Chemical structure of Safranin.

mesh. The chemical composition of the mineral used in this study isshown in Table 1.

2.2. Preparation of iron oxide–sepiolite composite

In brief, 2 g of Sep was added into a 100 mL solution of 1.55 gFeCl3·6H2O and 1.2 g FeSO4·7H2O and refluxed for 3 h in an oil bathat 110 °C under N2. When the mixed solution was cooled to 60 °C,~5mLNH4OH solution (25%)was added dropwise under N2with vigor-ous stirring to precipitate the surface-immobilized Fe2+ and Fe3+ ions.The pH of the final mixtures was controlled in the range of 10–11 to en-sure the complete transformation of iron ions to iron oxides. The mix-tures were aged at the same temperature for 2 h and then washedwith deionized water repeatedly. The obtained composite was dried at100 °C for 5 h.

X-ray diffraction (XRD) analysis of the natural sepiolite and MSepcomposite was performed using a PANalytical X'Pert PRO MPD instru-ment (PANalytical B.V., Almelo, The Netherlands) equipped with aback monochromator operating at a tube voltage of 40 kV and a tubecurrent of 30 mA using a copper cathode as the X-ray source (λ =1.542 Å). The 2θ angle that ranges from 5° to 80° was scanned with astep size of 0.02° and a time per step of 0.5 s.

Fourier transform infrared (FT-IR) spectra were recorded on aBruker tensor 27 spectrometer (Madison, WI, USA) using the standardKBr disk method (sample/KBr = 1/100). All transmittance spectrawere taken at the spectral resolution of 4 cm−1 for 64 scans over thewave number region 4000–500 cm−1.

Themorphology ofMSep compositewas also determinedwith scan-ning electronmicroscopy (SEM, KYKY-EM 3200, Zhongguancun Beijing,China), operated at 20 kV. The samples were sputter-coated with a thinlayer of gold for 2 min prior to SEM analysis.

A vibrating sample magnetometer (VSM, Lake Shore, 7410, USA)was applied to determine the magnetic property of the resultantcomposite.

The zeta potentials of materials were measured with a MalvernZetasizer (Nano-ZS, Malvern Instruments, Worcestershire, UK) accord-ing to literature [33]. In the preparation of the samples, 0.5 g of solidsample was mixed with 100 mL electrolyte (NaCl, 0.01 mol L−1) in anErlenmeyer flask and pH was adjusted using NaOH or HCl in the rangebetween 2.0 and 11.0. After stirring for 1 h at room temperature, thesamples were allowed to stand for 15 min to let larger particles settle.An aliquot taken from the supernatant was used to measure the zetapotential.

To determine the specific surface area of samples, Brunauer–Emmett–Teller (BET) measurements of N2 adsorption were carriedout at 77 K using a Belsorp-mini II (BEL Japan, Inc.). The samples weredried and out-gassed at 120 °C for a minimum of 6 h under vacuum be-fore the N2 adsorption experiments.

2.3. Adsorption studies

Batch experiments were carried out in 20 sets of 50 mL Erlenmeyerflasks with 10 mL dye solution agitated for 30 min using an incubatorshaker at 150 rpm. The experiments were performed under varying ini-tial concentrations (40–80 mg L−1), pH (6–10) and adsorbent dosage(0.03–0.07 g). The pH of the solution was adjusted using 0.1 mol L−1

KOH. After adsorption, theMSep composite was separated by a magnetand the sample absorbance was measured. The residual concentrationof the dye in the solution was then determined using the UV–vis

Table 1Main chemical composition of sepiolite sample.

Component (wt.%) SiO2 Al2O3 Fe2O3 CaO Na2O K2O MgO LOI

Sepiolite 53.9 0.21 0.01 2.94 0.01 0.01 24.22 18.7

677M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

spectrophotometer at 520 nm (λmax of the dye). Then the concentra-tions of the samples were determined by using linear regression equa-tion (A = 0.0826 × Ce − 0.0339, R2 = 0.9953) obtained by plotting acalibration curve for Safranin over a range of concentrations. The per-centage removal of Safranin was calculated using the following equa-tion:

%Removal ¼ Co � Ce

Co� 100 ð1Þ

where Co and Ce are the initial and final concentrations (mg L−1) ofthe dye in solution, respectively.

To establish the accuracy and reliability of the collected data, allsorption experiments were performed in triplicate and the mean ofthe three measures is taken for all calculations. The relative errors ofthe data were about 4%.

2.4. Experimental design and optimization

The RSMhas several classes of designs such as central composite de-sign (CCD), Doehlert matrix (DM) and Box–Behnken design (BBD)which are the most accepted designs applied by the researchers. Themost popular central composite design was employed in this work tostudy the variables for removal of Safranin from aqueous solution.This method has widely been used for fitting a second-order modeland it helps optimize the effective parameters with minimum experi-mental runs, as well as to analyze the interaction between the parame-ters and results [34]. For a 2 level study, the CCD consists of threeoperations namely: 2n factorial runs, 2n axial runs and nc center runs.For three independent variables, the design involved 8 factorial points,6 axial points and 6 replicates at the center points used for the purposeof estimating experimental error and the reproducibility of the data. Inthis research, the dosage of MSep (X1), initial concentration of Safranin(X2), and pH (X3) were selected as the set of three independent processvariables and the Safranin removal efficiency as a dependent output re-sponse variable. The selected process variables with their limits, unitsand notations are given in Table 2. Also, the structure of CCD designwith results of Safranin removal percent at different conditions isshown in Table 3. The designed experiments were carried in random-ized order to minimize the effects of the uncontrolled factors.

After conducting the required 20 tests and determining the final re-moval efficiency, the following second-order polynomial model equa-tion was developed to correlate the dependent and independentvariables.

Y ¼ β0 þ∑k

i¼1βiX j þ∑

k

i¼1βiiX

2i þ∑

k

i¼1∑k

j¼1βijXiX j þ ei ð2Þ

where Y is the predicted response (removal percentage), Xi and Xj

are independent variables, β0, βi, βii and βij are the constant, linear, qua-dratic and interaction coefficients, respectively, and ei is the randomerror. Analysis of variance (ANOVA) was used to estimate the signifi-cance of the main effects and interactions. The experimental design

Table 2Experimental range and levels of independent process variables.

Variables Factors Ranges and levels

xi −α −1 0 +1 +α

Adsorbent dosage (g) X1 0.016 0.03 0.05 0.07 0.083Initial dye concentration(mg L−1)

X2 26.3 40 60 80 93.6

pH X3 4.6 6 8 10 11.3

data were analyzed using Design-Expert (version 8.0.7.1, STAT-EASEInc., Minneapolis, MN, USA) software.

3. Results and discussion

3.1. Characterization of the composites

The XRD pattern of natural sepiolite is shown in Fig. 2(a). Thereflections indexed to the (110), (130), (060), (131), (260), (241),(080), (331), (341), (441), (371), (202), (541) and (791) planesappearing at 2θ = 7.71, 12.03, 19.96, 20.83, 24.03, 25.43, 26.82,28.19, 29.56, 33.51, 35.23, 36.85, 40.08 and 60.10° respectively, areconsistent with the standard XRD data for the orthorhombic unit-cell of sepiolite based on comparison with the standard pattern of se-piolite (JCPDS card No. 13-0595). Besides, the other peaks can beindexed to 44.01° (202), 58.22° (122), 66.28° (300), and 72.95°(128) as lattice planes of calcite (JCPDS card No. 05-0586). In Fig.2(b), the characteristic reflections of magnetite (Fe3O4) are presentin synthesized Fe3O4@sepiolite nanocomposite, which can beindexed to 30.46° (220), 35.84° (311), 43.61° (400), 54.09° (422),57.59° (511), and 62.99° (440) as planes of a typical of spinel phasesof iron oxide according to JCPDS 75-0449.

FTIR spectra of natural sepiolite, Fe3O4, andMSepnanocomposite areshown in Fig. 3. Absorption bands appearing at 3567 cm−1, 3418 cm−1

and 1662 cm−1 in the FTIR spectra are attributed to diverse water mol-ecules existing in the mineral [35]. The presence of characteristic ab-sorption bands at 1079 cm−1, and 691 cm−1, suggests the existenceof CaCO3 [36], which is related to calcite in the natural sepiolite mineral.The bands at 1079 cm−1 and 691 cm−1 are assigned to symmetricstretching [37] and in-plane bending [38] of CO3

2-, respectively. Thecharacteristic bands of Si–O–Si can be observed at 1016 cm−1,1209 cm−1 and 974 cm−1 [39]. The band at 646 cm−1 represents thebending vibration of the Mg–OH bond of both samples [40]. The broadband at 3442 cm−1 forMSep sample indicates themonomeric hydrogenbond vibrations. The band centered at 1091 cm−1 for MSep can be at-tributed to the presence of Si–O–Fe bond. In addition, the absorptionband around 563 cm−1 is the characteristic band of Fe3O4, correspond-ing to Fe–O vibrations of Fe3O4 [41]. This band is also found in the FTIRspectrum ofMSep, compatible with the presence of Fe3O4 nanoparticlesin the prepared composite.

The SEM micrographs of the natural sepiolite and MSep nanocom-posite are presented in Fig. 4(a) and (b). The SEM image of the preparedMSep nanocomposite shows the presence of sepiolite fibers surroundedby iron oxide nanoparticles. It reveals that the surface coverage of mag-netic nanoparticles on sepiolite is high. The nanoparticles exhibit abroad distributionwith sizes ranging from 29 nm to 53 nm. The averageparticle diameter, based on measurement of 50 nanoparticles, wasfound to be 41 ± 2.6 nm.

The magnetic hysteresis loop of MSep nanocomposite at 298 K is il-lustrated in Fig. 4(c). The saturation magnetization was found to be34.2 emu g−1. The saturation magnetization value was more thanother literature value (31.8 emu g−1) reported for Msep preparation[42]. This differencemay be attributed to the effect of the proposed syn-thesis method. The aqueous dispersion of the MSep material fast accu-mulates near the vessels upon using a magnet, leaving the solutionclear and transparent (inset of Fig. 4c).

The zeta potentials of MSep and natural Sep at different pH valueswere measured, and the results are shown in Fig. 4(d). It was foundthat the isoelectric points of natural Sep and MSep were about 6.8 and7.9, respectively. Different values, ranging from pH 3.2 to 7.4 havebeen reported [35,43] for the isoelectric point of natural Sep, and 7.7to 8.5 for that of iron-coated Sep [42,44]. This phenomenon can beassigned to the purity and the mineralogical composition of sepioliteused in the different studies, as well as to the preparing method of themineral for the electrokinetic measurements [43].

Table 3Experimental design matrix and results.

Run Coded level of variables Actual level of variables Y(% of dye removal)

X1 X2 X3 Adsorbent dosage, X1 (g) Initial dye concentration, X2 (mg L−1) pH, X3 Experiment Predicted

1 −1 −1 −1 0.030 40.00 6.00 65.50 66.112 +1 −1 −1 0.070 40.00 6.00 81.72 82.333 −1 +1 −1 0.030 80.00 6.00 48.14 47.834 +1 +1 −1 0.070 80.00 6.00 73.37 73.715 −1 −1 +1 0.030 40.00 10.00 80.83 77.616 +1 −1 +1 0.070 40.00 10.00 92.25 93.837 −1 +1 +1 0.030 80.00 10.00 64.65 59.338 +1 +1 +1 0.070 80.00 10.00 86.37 85.209 −1.682 0 0 0.016 60.00 8.00 42.70 46.7610 +1.682 0 0 0.083 60.00 8.00 83.81 82.1611 0 −1.682 0 0.050 26.36 8.00 87.54 86.9512 0 +1.682 0 0.050 93.63 8.00 61.33 64.3313 0 0 −1.682 0.050 60.00 4.63 74.72 71.2814 0 0 +1.682 0.050 60.00 11.36 88.47 90.6215 0 0 0 0.050 60.00 8.00 80.37 80.9516 0 0 0 0.050 60.00 8.00 80.15 80.9517 0 0 0 0.050 60.00 8.00 80.46 80.9518 0 0 0 0.050 60.00 8.00 80.91 80.9519 0 0 0 0.050 60.00 8.00 80.29 80.9520 0 0 0 0.050 60.00 8.00 80.25 80.95

678 M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

The specific surface areas and the pore size distribution of natural se-piolite and magnetic sepiolite were measured from the N2 adsorption/desorption isotherms using the BET method. Based on BET theory, thespecific surface area was calculated as 137.54 m2 g−1 for Sep and138.21 m2 g−1 for MSep. The nitrogen adsorption/desorption isothermof natural Sep andMSep is shown in Fig. 5. The type IV isotherm indicat-ed the existence of mesopores in both materials [45]. According to theIUPAC classification of pore dimensions, there are three categoriesgrouped as micropores (diameter, d ≤ 2 nm), mesopores(2 b d b 50 nm) and macropores (d ≥ 50 nm) [46]. Maximum pore are

Fig. 2. XRD patterns of (a) natural sepio

mesopores followed by a small fraction ofmicropores. The pore size dis-tribution was calculated according to the BJHmethod. The average porediameter of natural Sep and MSep was mainly distributed within therange of 2–50 nm and partly more than 50 nm, which suggests thatboth Sep and MSep comprised mesopores and macropores. Further-more, the average pore diameter and pore volume of Sep were10.93 nm and 0.376 cm3 g−1, which increased to 16.52 nm and0.571 cm3 g−1 in MSep. The pore volume of Sep increased from 0.376to 0.571 (~52%) after Fe3O4 impregnation, mainly due to bigger surfacearea of Fe3O4 nanoparticle in comparison to natural clay particle.

lite and (b) MSep nanocomposite.

Fig. 3. FTIR spectra of (a) Fe3O4, (b) MSep nanocomposite and (c) natural sepiolite.

679M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

3.2. Response surface modeling

3.2.1. Development of regression model equationIn the present study, a polynomial regression modeling was per-

formed between the predicted values of percent removal (Y) and thecorresponding coded values (X1, X2 and X3) of the three independentvariables (MSep dosage, initial concentration of Safranin, and pH), andsubsequently a predictive model as follows was obtained, as describedin Eq. (3).

Y %ð Þ ¼ 80:95þ 10:52X1 � 6:73X2 þ 5:75X3 þ 2:41X1X2 � 5:83X21

� 1:88X22 ð3Þ

The positive sign in front of the termsmeans that the correspondingterms affect the response positively and negative sign affects it nega-tively. Based on the experimental design results, it has been foundthat the selected variables have significant effect on the removal effi-ciency of Safranin from 46.76% to 93.83%.

3.2.2. Analysis of variance (ANOVA) analysisThe adequacy of themodel was evaluated by the sumof squares, DF,

mean square, F-values and p-values for ANOVA, as listed in Table 4. Ac-cording to Table 4, the sum of square is 451.16 for pH, 617.74 for dyeconcentration, and 1512.64 for adsorbent dosage. This means that theeffect of adsorbent dosage was more important than pH and dye

Fig. 4. SEM image of (a) natural sepiolite and (b) MSep nanocomposite; (c) Magnetization curve for the MSep nanocomposite at room temperature (inset shows the nanocomposite re-sponse to a magnet); (d) Zeta potential of natural sepiolite and MSep nanocomposite versus pH.

680 M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

concentration. TheANOVA results of the second-order regressionmodel(Eq. (3)) suggest that the model is statistically significant, as evidentfrom the Fisher's variance ratio test (F-test, Fmodel = 48.39) with avery low probability value (pmodel b 0.0001). Moreover, the high valueof the coefficient of determination (R2= 0.978) between the experi-mental and predicted removal efficiencies was obtained. This impliesthat regressionmodel provides an excellent explanation of the relation-ship between the independent variables (MSep dosage, initial concen-tration of Safranin, and pH) and 97.8% of the variations of dye removalefficiency.

3.2.3. Three-dimensional (3D) response surfacesThe three-dimensional graphical representations of the regression

equation facilitate an examination of the individual and interactive ef-fects of the test variables on the response. The effects of three indepen-dent variables on the dye removal efficiency are shown in the 3D surfacegraphs (Fig. 6). It was found that dye removal efficiency is highly depen-dent on the initial Safranin concentration (Fig. 6(a)). The Safraninremoval efficiency decreased with the increase of the initial dyeconcentration and increased with the increase of MSep dosage. Athigher initial Safranin concentration, the largest number of Safranin

molecules would quickly saturate all the binding sites of MSepnanocomposite. This could be ascribed to the accompanying de-crease in dye removal efficiency due to the limited accessible sitesat an initial Safranin concentration of 80 mg L−1. Also, increasing theadsorbent dosage from 0.03 to 0.07 g leads to better Safranin uptake,so that the number of available binding and reactive sites increaseswith rising MSep dosage.

According to Fig. 6(b), the effectiveness of MSep in removal of theSafranin dye is also affected by solution pH. The adsorption of Safraningradually increases with increasing the initial pH of dye-containing so-lutions from 6.0 to 10.0. It is well known that the pH of the dye solutioninfluences the surface charge of the adsorbent materials, the dissocia-tion of functional groups present on the adsorbent as well as the degreeof ionization of the adsorbate. The MSep nanocomposite surface is pos-itively charged when pH b 7.9 due to protonation of surface functionalgroups and is negatively charged when pH N 7.9 related to dehydroxyl-ation or deprotonation of the MSep surface groups.

At lower pH values, the hydrogen ions compete with the dye mole-cules for the available sorption sites, thus the removal efficiency de-creases. With the increase in pH of the solution, increase of theremoval efficiency can be explained by the electrostatic attraction

Fig. 5. N2 adsorption–desorption isotherms and pore size distributions of (a) natural sepi-olite and (b) MSep nanocomposite.

681M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

between the negatively charged surface of the MSep nanocompositeand cationic Safranin molecules.

The effect of pH and dosage of MSep on the Safranin removal effi-ciency is presented in Fig. 6(c). The results showed that the removal ef-ficiency is increased by risingMSep dosage at a constant pH.Whenever,the dosage of MSep is constant, the dye removal efficiency is increasedby increasing the solution pH. The reason is that at higher pH, the sur-face charge ofMSep becomesmore negatively charged, which enhancesSafranin adsorption through electrostatic attraction.

Table 4ANOVA results for the response surface quadratic model for Safranin removal.

Source Sum of squares D.F.a Mean squares F-value Prob N F Remarks

Model 3167.37 9 351.93 48.39 b0.0001 SignificantX1 1512.64 1 1512.64 207.98 b0.0001 SignificantX2 617.74 1 617.74 84.93 b0.0001 SignificantX3 451.16 1 451.16 62.03 b0.0001 SignificantX1X2 46.61 1 46.61 6.41 0.0298 SignificantX1X3 8.63 1 8.63 1.19 0.3015X2X3 1.67 1 1.67 0.23 0.6426X12 477.63 1 477.63 65.67 b0.0001 Significant

X22 46.91 1 46.91 6.45 0.0294 Significant

X32 7.62 1 7.62 1.05 0.3302

Residual 72.73 10 7.27Lack of fit 72.37 5 14.47 200.17 b0.0001Pure error 0.36 5 0.072Cor. total 3240.10 19

a D.F.— degrees of freedom.

3.3. Effect of shaking time

The effect of shaking time on the amount of Safranin adsorbed onMSep was investigated in the range of 0–90 min at 25 °C using the op-timum amount of adsorbent (0.06 g) for 10 mL of 40 mg L−1 Safraninand the results are shown in Fig. 7. As can be seen the dye adsorptionproceeds through a two-stage process, involving a rapid initial adsorp-tion of Safranin on the MSep surface followed by a second stage witha much slower adsorption. It is evident that most of the Safranin is re-moved within 10 min at which the adsorption amount (qt) is~10.7 mg Safranin/g of MSep, and almost levels off after this period, sothat the Safranin uptake remains almost unchanged with increasingthe contact time. The fast initial removal rate is probably due to thefast diffusion of Safranin from the solution onto the external surface ofMSep composite. As the sites are gradually occupied, the adsorbed Saf-ranin tends to be transported from the bulk phase to the actual sorptionsites (i.e., inner-sphere pores of MSep). Such a slow diffusion processwill decrease the sorption rate of Safranin at later stages. After 30 minof continuous shaking, a quasi-steady-state approximation can be con-sidered. Accordingly in all the batch experiments a contact time of30 min was conducted under vigorous shaking.

3.4. Adsorption kinetic

The kinetic model of solute adsorption at solid/solution interfaces isusually complex. The adsorption rate is highly dependent on several pa-rameters such as the status of the solidmatrix that has generally hetero-geneous reactive sites, and the physicochemical conditions ofadsorption. Two of the most widely used kinetic models, i.e. pseudo-first-order [47] equation and pseudo-second-order [48] equation wereused to research the adsorption kinetic behavior of Safranin ontoMSep. The consistency between the experimental and the model-pre-dicted data was investigated by calculating correlation coefficients (R2

values closer to 1 means more applicability of the model) and by ob-serving the extent to which the experimental adsorption capacity isclose to the theoretical value. Lagergren pseudo-first-order model iscommonly expressed as follows:

ln qe � qtð Þ ¼ lnqe � k1t ð4Þ

where qe and qt (mg g−1) are the adsorption capacities at equilibri-um and at time t, respectively, and k1 is the rate constant of the pseudo-first-order adsorption (min−1). Using this well-known equation, thevalues of k1 and qe were calculated from the slope and intercept of theplot of ln(qe-qt) versus t (Fig. 8(a)), respectively [49]. The correspond-ing kinetic parameters from both models are listed in Table 5.

The pseudo-second-order model equation is expressed by the fol-lowing equation:

tqt

¼ 1k2q2e

þ 1qe

t ð5Þ

where k2 is the equilibrium rate constant of pseudo-second-order ad-sorption (gmg−1min−1). The slope and intercept of the plot of t/qt ver-sus t were used to calculate the second-order rate constant, k2 (Fig.8(b)). From Table 5, the pseudo first- and pseudo second-order modelsfit the kinetic data with adequate accuracy (R2 N 0.876). However, thecalculated adsorption capacity of the pseudo-second order reaction(10.8 mg g−1) and experimental qe (10.7 mg g−1) has very similarvalues. It implies that the adsorption process may be the rate-limitingstep involving valence forces through sharing or exchange of electronsbetween the adsorbent and the adsorbate [50].

Fig. 6. Effects of considered factors on Safranin removal efficiency.

682M.Fayazietal./JournalofM

olecularLiquids

212(2015)

675–685

Fig. 7. Effect of the shaking time on Safranin adsorption capacity ontoMSep nanocompos-ite. The inset shows Weber–Morris intra-particle diffusion plot of qt versus t1/2.

Table 5Kinetic parameters for Safranin adsorption on the MSep nanocomposite.

Pseudo-first-order Pseudo-second-order

qe,cal (mgg−1)

qe,exp (mgg−1)

k1(min−1)

R2 qe,cal (mgg−1)

k2 (g mg−1

min−1)R2

2.53 10.7 0.215 0.876 10.8 0.193 0.999

683M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

3.5. Adsorption isotherm studies

Isotherm studies can describe how the adsorbates interact with ad-sorbents, affording the most important parameter for designing a de-sired adsorption system. The adsorption isotherms of Safranin on theMSep composite at different initial concentrations are given in Fig. 9,

Fig. 8.Pseudo-first-order kinetics (a) andpseudo-second-order kinetics (b) of Safranin ad-sorption on the MSep nanocomposite.

and the equilibrium adsorption data were analyzed by the well-known Langmuir and Freundlich isotherm models. The Langmuir iso-therm assumes that monolayer adsorption occurs at binding sites withhomogenous energy levels, no interactions between adsorbed mole-cules and no transmigration of adsorbed molecules on the adsorptionsurface [51]. The Langmuir equations can be expressed as:

qe ¼qmklCe

1þ klCeð6Þ

or

Ce

qe¼ 1

klqmþ 1qm

Ce ð7Þ

where Ce is the equilibrium concentration of the Safranin solution(mg L−1), qe is the adsorption capacity at equilibrium (mg g−1), kl isthe constant related to free energy of adsorption (L mg−1), and qm isthe maximum adsorption capacity at monolayer coverage (mg g−1).

Fig. 9. Experimental adsorption data fitted to Langmuir (a) and Freundlich (b) models(temperature: 25 °C; contact time: 30 min).

Fig. 10. (A) Van't Hoff plots for the uptake of Safranin on the MSep nanocomposite.

684 M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

The Freundlich isotherm is an empirical equation that assumes het-erogeneous adsorbent surface with its adsorption sites at varying ener-gy levels [52]. The corresponding equations are commonly representedby:

qe ¼ kf C1=ne ð8Þ

or

lnqe ¼ lnkf þ1n

lnCe: ð9Þ

kf (mg g−1 (L mg−1)1/n) and n are the Freundlich constants charac-teristics of the system, indicating the adsorption capacity and the ad-sorption intensity, respectively. If the value of 1/n is lower than 1, itindicates a normal Langmuir isotherm; otherwise, it is indicative of co-operative adsorption [53].

The Langmuir and Freundlich constants and the calculated coeffi-cients are listed in Table 6. The adsorption of Safranin was well fittedto the Langmuir isotherm model with the higher R2 (0.98). It indicatedthe adsorption took place at specific homogeneous sites within the ad-sorbent forming monolayer coverage of Safranin on the surface of theabsorbent. The Freundlich constant 1/n was smaller than 1, indicatinga favorable process. Moreover, the essential feature of the Langmuir iso-therm can be expressed in terms of a dimensionless constant separationfactor (RL) given by the following equation [54]:

RL ¼ 11þ klC0

ð10Þ

where kl (L mg−1) is the Langmuir constant and C0 (mg L−1) is the ini-tial concentration in the liquid phase. The value of RL indicates the shapeof the isotherm to be either unfavorable (RL N 1), linear (RL = 1), favor-able (0 b RL b 1) or irreversible (RL=0) [55]. In the study, the value of RL

calculated for the initial concentrations of Safranin was 0.21. Since theresult is within the range of 0–1, the adsorption of Safranin onto the ad-sorbent appears to be a favorable process.

3.6. Thermodynamic studies

Thermodynamic parameters such as the standard Gibbs free energyof adsorption ΔG0, standard entropy of adsorption (ΔS0) and standardenthalpy of adsorption (ΔH0) provide additional information on inher-ent energetic changes of adsorption process, which can be calculatedby using Eqs. (11) and (12):

ΔG0 ¼ �RT lnkd ð11Þ

lnkd ¼ ΔS0

R� ΔH0

RTð12Þ

where kd is the distribution coefficient (kd=qe/Ce), T is the temperature,and R is the gas constant (8.314 J mol−1 K−1). The values of (ΔH0) and(ΔS0) were calculated from the slope and intercept of the Van't Hoff lin-ear plot of ln (kd) against 1/T (Fig. 10). It can be seen in Table 7 that allvalues obtained for ΔG0 are negative, suggesting the spontaneous na-ture the Safranin adsorption byMSep nanocomposite. The observed de-crease in negative values of ΔG0 with increasing temperature impliedthat the adsorption became less favorable at higher temperatures. The

Table 6Isotherm parameters for Safranin adsorption on the MSep nanocomposite.

Langmuir isotherm Freundlich isotherm

kl (L mg−1) qm (mg g−1) RL R2 kf (mg g−1 (L mg−1)1/n) n R2

0.0945 18.48 0.21 0.989 2.34 1.83 0.958

change in Gibbs energy for physisorption is between −20 and0 kJ mol−1, but chemisorption is in a range of −80 to −400 kJ mol−1

[56]. The values of ΔG0 obtained in this study are within the ranges of−20 and 0 kJ mol−1, indicating that physisorption is the dominatingmechanism. On the other hand, the observed positive value of ΔH0

(Table 7) indicates that the process is endothermic, while the positivevalue of ΔS0 corresponded to an increase in randomness at the solid/so-lution interface during the adsorption of Safranin by MSep.

3.7. Adsorption mechanism

The proposition of mechanism in the adsorption study is still theforemost challenge. A possible adsorption mechanism between targetdye and functional groups on the adsorbent has been proposed asshown in Fig. 11. As can be seen, electrostatic attraction and hydrogenbonds are found to be the dominant interactions between Safraninand adsorbent surface. At higher solution pH values, the surface of sepi-olite particles may become negatively charged, which enhances thepositively charged Safranin cations through electrostatic forces of at-traction. In this case, it can be written as follows:

SO− þ Dyeþ ¼ SO−⋯Dyeþ: ð13Þ

As a result, Safranin ions are thus more favorably adsorbed onto thesurfaces of the adsorbent at relative higher pH values. Hence, Safraninremoval by MSep is mainly described through physisorption with aminor contribution from chemisorption mechanisms. A similar mecha-nism has been previously reported by Alkan et al. [57] for maxilon blue5G adsorption on sepiolite particles.

4. Conclusions

In this study,MSepnanocomposite adsorbentmaterialwaspreparedvia chemical co-precipitation in the air and characterized by XRD, FTIR,SEM and BET analyses. The resulting composite combined both featuresof Fe3O4 and sepiolite, and thus exhibited extraordinary adsorptioncapacity and fast removal rate for Safranin dye. RSM was employed byutilizing the CCD to optimize important parameters including the

Table 7Thermodynamic parameters at different temperatures.

ΔH (kJ mol−1) ΔS (J mol−1 K−1) ΔG (kJ mol−1)

298 K 308 K 318 K

2.63 56.19 −14.12 −14.68 −15.24

Fig. 11. Possible Safranin adsorption mechanism on the adsorbent structure.

685M. Fayazi et al. / Journal of Molecular Liquids 212 (2015) 675–685

solution pH, initial dye ion concentration and absorbent dosage. Qua-dratic model was developed to correlate the process variables and theresponse. The experimental values obtained for the Safranin removal ef-ficiency were found to agree satisfactorily with the values predicted bythe proposed model with a high coefficient of determination (R2 =0.978). Moreover, the adsorption isotherms and kinetics were investi-gated and indicate that the equilibrium and kinetic adsorption werewell-modeled by the Langmuir isotherm model and the pseudo-sec-ond-order kinetic model, respectively. The calculated thermodynamicparameters suggested that the adsorption process is thermodynamical-ly favorable, spontaneous, and endothermic in nature. Electrostatic at-traction and hydrogen bonds are found to be the dominantinteractions between Safranin and adsorbent surface. Thereby, it canbe concluded that the MSep nanocomposite can efficiently be utilizedas a magnetically separable and efficient adsorbent for dyeing effluenttreatment.

References

[1] H. Zollinger, Color Chemistry: Syntheses, Properties, and Applications of OrganicDyes and Pigments, VCH, 1991.

[2] M. Ghaedi, M. Pakniat, Z. Mahmoudi, S. Hajati, R. Sahraei, A. Daneshfar, Spectrochim.Acta A 123 (2014) 402–409.

[3] A. Ebadi, A.A. Rafati, J. Mol. Liq. 209 (2015) 239–245.[4] S. Shariati, M. Faraji, Y. Yamini, A.A. Rajabi, Desalination 270 (2011) 160–165.[5] Q. Sun, L. Yang, Water Res. 37 (2003) 1535–1544.[6] V.K. Gupta, A. Mittal, R. Jain, M. Mathur, S. Sikarwar, J. Colloid Interface Sci. 303

(2006) 80–86.[7] V.K. Gupta, R. Jain, A. Mittal, M. Mathur, S. Sikarwar, J. Colloid Interface Sci. 309

(2007) 464–469.[8] S. Chowdhury, P. Saha, CLEAN – Soil, Air, Water 39 (2011) 274–282.[9] F. Nasiri Azad, M. Ghaedi, K. Dashtian, S. Hajati, A. Goudarzi, M. Jamshidi, New J.

Chem. (2015).[10] S. Preethi, A. Sivasamy, S. Sivanesan, V. Ramamurthi, G. Swaminathan, Ind. Eng.

Chem. Res. 45 (2006) 7627–7632.[11] A. Dąbrowski, Adv. Colloid Interf. Sci. 93 (2001) 135–224.[12] A.K. Golder, N. Hridaya, A.N. Samanta, S. Ray, J. Hazard. Mater. 127 (2005) 134–140.[13] H. Sadegh, R. Shahryari-ghoshekandi, S. Agarwal, I. Tyagi, M. Asif, V.K. Gupta, J. Mol.

Liq. 206 (2015) 151–158.[14] S.-Y. Wang, M.-H. Tsai, S.-F. Lo, M.-J. Tsai, Bioresour. Technol. 99 (2008) 7027–7033.[15] T. Motsi, N.A. Rowson, M.J.H. Simmons, Int. J. Miner. Process. 92 (2009) 42–48.[16] A.E.-A.A. Said, A.G. Ludwick, H.A. Aglan, Bioresour. Technol. 100 (2009) 2219–2222.[17] G. Seeber, P. Brunner, M.R. Buchmeiser, G.K. Bonn, J. Chromatogr. A 848 (1999)

193–202.[18] S.R. Shirsath, A.P. Hage, M. Zhou, S.H. Sonawane, M. Ashokkumar, Desalination 281

(2011) 429–437.[19] C. Namasivayam, D. Prabha, M. Kumutha, Bioresour. Technol. 64 (1998) 77–79.[20] A. Esteban-Cubillo, J.F. Marco, J.S. Moya, C. Pecharromán, J. Phys. Chem. C 112 (2008)

2864–2871.

[21] S. Lazarević, Ž. Radovanović, D. Veljović, A. Onjia, D. Janaćković, R. Petrović, Appl.Clay Sci. 43 (2009) 41–48.

[22] M. Alkan, Ö. Demirbaş, S. Çelikçapa, M. Doğan, J. Hazard. Mater. 116 (2004)135–145.

[23] E. González-Pradas, M. Villafranca-Sánchez, M. Socías-Viciana, M. Fernández-Pérez,M.D. Ureña-Amate, J. Chem. Technol. Biotechnol. 74 (1999) 417–422.

[24] A. Sheikhhosseini, M. Shirvani, H. Shariatmadari, Geoderma 192 (2013) 249–253.[25] A. Afkhami, R. Norooz-Asl, Colloids Surf. A 346 (2009) 52–57.[26] T. Madrakian, A. Afkhami, M. Ahmadi, H. Bagheri, J. Hazard. Mater. 196 (2011)

109–114.[27] I.-H. Cho, K.-D. Zoh, Dyes Pigments 75 (2007) 533–543.[28] J.-P. Wang, Y.-Z. Chen, X.-W. Ge, H.-Q. Yu, Colloids Surf. A 302 (2007) 204–210.[29] A.M. Ghaedi, M. Ghaedi, A. Vafaei, N. Iravani, M. Keshavarz, M. Rad, I. Tyagi, S.

Agarwal, V.K. Gupta, J. Mol. Liq. 206 (2015) 195–206.[30] A.A. Ahmad, B.H. Hameed, A.L. Ahmad, J. Hazard. Mater. 170 (2009) 612–619.[31] A.A. Ahmad, B.H. Hameed, J. Hazard. Mater. 173 (2010) 487–493.[32] S. Bakhtiary, M. Shirvani, H. Shariatmadari, Microporous Mesoporous Mater. 168

(2013) 30–36.[33] S. Dikmen, G. Yilmaz, E. Yorukogullari, E. Korkmaz, Can. J. Chem. Eng. 90 (2012)

785–792.[34] I.A.W. Tan, A.L. Ahmad, B.H. Hameed, J. Hazard. Mater. 153 (2008) 709–717.[35] S. Lazarević, I. Janković-Častvan, D. Jovanović, S. Milonjić, D. Janaćković, R. Petrović,

Appl. Clay Sci. 37 (2007) 47–57.[36] M. Tatzber, M. Stemmer, H. Spiegel, C. Katzlberger, G. Haberhauer, M.H. Gerzabek,

Environ. Chem. Lett. 5 (2007) 9–12.[37] J. Ihli, Y.-Y. Kim, E.H. Noel, F.C. Meldrum, Adv. Funct. Mater. 23 (2013) 1575–1585.[38] S. Gopi, V.K. Subramanian, K. Palanisamy, Mater. Res. Bull. 48 (2013) 1906–1912.[39] E. Eren, H. Gumus, N. Ozbay, Desalination 262 (2010) 43–49.[40] A. Tabak, E. Eren, B. Afsin, B. Caglar, J. Hazard. Mater. 161 (2009) 1087–1094.[41] V. Chandra, J. Park, Y. Chun, J.W. Lee, I.-C. Hwang, K.S. Kim, ACS Nano 4 (2010)

3979–3986.[42] H. Liu, W. Chen, C. Liu, Y. Liu, C. Dong, Microporous Mesoporous Mater. 194 (2014)

72–78.[43] E. Sabah, U. Mart, M. Çınar, M.S. Çelik, Sep. Sci. Technol. 42 (2007) 2275–2288.[44] S. Lazarević, I. Janković-Častvan, V. Djokić, Z. Radovanović, D. Janaćković, R. Petrović,

J. Chem. Eng. Data 55 (2010) 5681–5689.[45] J. Lin, Y. Zhan, Chem. Eng. J. 200–202 (2012) 202–213.[46] R.A. Reza, M. Ahmaruzzaman, RSC Adv. 5 (2015) 10575–10586.[47] I. Langmuir, J. Am. Chem. Soc. 40 (1918) 1361–1403.[48] E. Bulut, M. Özacar, İ.A. Şengil, J. Hazard. Mater. 154 (2008) 613–622.[49] J.C. Bellot, J.S. Condoret, Process Biochem. 28 (1993) 365–376.[50] L. Li, L. Fan, H. Duan, X. Wang, C. Luo, RSC Adv. 4 (2014) 37114–37121.[51] M.A.K.M. Hanafiah, W.S.W. Ngah, S.H. Zolkafly, L.C. Teong, Z.A.A. Majid, J. Environ.

Sci. 24 (2012) 261–268.[52] H. Freundlich, Über die Adsorption in Lösungen, W. Engelmann, 1906.[53] L. Fan, Y. Zhang, X. Li, C. Luo, F. Lu, H. Qiu, Colloids Surf. B 91 (2012) 250–257.[54] L. Ai, C. Zhang, Z. Chen, J. Hazard. Mater. 192 (2011) 1515–1524.[55] G. McKay, J. Chem. Technol. Biotechnol. 32 (1982) 759–772.[56] N.M. Mahmoodi, B. Hayati, M. Arami, J. Chem. Eng. Data 55 (2010) 4638–4649.[57] M. Alkan, M. Doğan, Y. Turhan, Ö. Demirbaş, P. Turan, Chem. Eng. J. 139 (2008)

213–223.