Ethylenediamine grafted to graphene oxide@Fe3O4 for chromium(VI) decontamination: Performance, modelling, and fractional factorial design

A method for grafting ethylenediamine to a magnetic graphene oxide composite (EDA-GO@Fe3O4) was developed for Cr(VI) decontamination. The physicochemical properties of EDA-GO@Fe3O4 were characterized using HRTEM, EDS, FT-IR, TG-DSC, and XPS. The effects of pH, sorbent dose, foreign anions, time, Cr(VI) concentration, and temperature on decontamination process were studied. The solution pH can largely affect the decontamination process. The pseudo-second-order model is suitable for being applied to fit the adsorption processes of Cr(VI) with GO@Fe3O4 and EDA-GO@Fe3O4. The intra-particle diffusion is not the rate-controlling step. Isotherm experimental data can be described using the Freundlich model. The effects of multiple factors on the Cr(VI) decontamination was investigated by a 25−1 fractional factorial design (FFD). The adsorption process can significantly be affected by the main effects of A (pH), B (Cr(VI) concentration), and E (Adsorbent dose). The combined factors of AB (pH × Cr(VI) concentration), AE (pH × Adsorbent dose), and BC (Cr(VI) concentration × Temperature) had larger effects than other factors on Cr(VI) removal. These results indicated that EDA-GO@Fe3O4 is a potential and suitable candidate for treatment of heavy metal wastewater.


Introduction
Heavy metal pollution is a current worldwide environmental concern because it can harm ecosystems and endanger human health. Since the industrial revolution, chromium has been widely used in electroplating, tanning, dying, smelting, and corrosion protection [1][2][3]. Cr (VI), one form of chromium, is very harmful to most organisms due to its mammalian toxicity PLOS  and carcinogenicity [4]. Therefore, it is necessary and important to separate Cr(VI) ions from aqueous solution before they are discharged into aquatic systems. Compared with traditional chemical precipitation and ion exchange methods, adsorption is a simpler, faster, and more economically viable method for removing heavy metals from various wastewaters [5][6][7]. The nature of an adsorbent is critical to the adsorption process, and the efficiency and cost of which are determined by the efficiency of the adsorbent regarding contaminant removal and its solid-liquid separation ability [1]. Therefore, it is desirable to find adsorbents that possess both high adsorption ability and straightforward solid-liquid separation.
In recent years, graphene oxide (GO) has been used as an excellent adsorbent material due to its unique properties [8,9]. GO has very high surface area and a large number of carboxyl, hydroxyl, carbonyl, and epoxy groups [8,10], which can be used as anchoring sites for metal ions. GO and GO-based materials have been used as adsorbents for binding metals such as chromium [11], cadmium [12], lead [13], zinc [14], platinum [15], and copper [16]. However, due to its nanoscale and hydrophilic nature, GO is difficult to separate from aqueous solution following the adsorption process. The dispersion of magnetic nanomaterials on GO sheets is a topic of current research because it combines the advantages of high adsorption rate and easy phase separation [17,18]. Therefore, it is important to integrate graphene oxide with magnetic nanomaterials for improving the solid-liquid separation capacity of the composite [19]. The adsorption capacity of an adsorbent for contaminants is partly determined by the number of functional groups [20]. Ethylenediamine is low-toxicity and low-cost, and contains two amino groups that can form stable chelates with metal ions. Therefore, grafting ethylenediamine to GO and GO-based materials may increase their adsorption ability. However, the adsorption behaviors of ethylenediamine modified magnetic graphene oxide composite (EDA-GO@-Fe 3 O 4 ) for Cr(VI) ions have not been fully investigated.
It is well known that environmental factors including pH, contact time, temperature, initial metal concentration, and background electrolyte species may affect the efficiency of an adsorbent for metal ions, and that this could be increased by optimizing these factors [21]. Traditional one-factor experimental design just study one factor at a time, which cannot investigate the interaction of factors [22]. Full factorial experimental design can give information about the interaction of factors, but it is only suitable for experiments with a small number of factors [21]. Fractional factorial design (FFD) can identify significant factors and assess interaction of factors only with a smaller number of experiments [21,23]. Besides, the produced results can be easily analyzed without any complicated calculations. Therefore, it is significant to identify the key factors that have large effects on the Cr(VI) decontamination by EDA-GO@Fe 3 O 4 using FFD.
In this study, a novel type of GO based composite named EDA-GO@Fe 3 O 4 was developed for effective Cr(VI) decontamination. To the authors' knowledge, few studies attempted to graft ethylenediamine to magnetic graphene oxide for improving the Cr(VI) removal efficiency. Moreover, there has not been any studies on the use of FFD to identify the main factors influencing adsorption efficiency of EDA-GO@Fe 3 O 4 for Cr(VI) ions. The aims of this research are to: (1) synthesize and characterize magnetic hybrid adsorbent (EDA-GO@Fe 3 O 4 ) and apply it for removing Cr(VI) ions from wastewater; (2) evaluate the effects of pH, sorbent dose, foreign anions, time, Cr(VI) concentration, and temperature on removal process; (3) investigate the reusability of EDA-GO@Fe 3 O 4 composite; (4) apply kinetics and isotherm models for modelling the adsorption experiments; and (5) use FFD to identify significant factors and interactions for removing Cr(VI) ions with EDA-GO@Fe 3 O 4 .

Synthesis of EDA-GO@Fe 3 O 4
The GO was prepared using the modified Hummers procedure reported previously [17]. Natural graphite was first preoxidized with K 2 S 2 O 8 , P 2 O 5 , and H 2 SO 4 , then further oxidized with H 2 SO 4 , NaNO 3 , and KMnO 4 . Lastly, the graphite oxide layers were separated by ultrasonication to obtain a GO suspension. Coprecipitation method was used to synthesize the magnetic graphene oxide (GO@Fe 3 O 4 ). Fe 2+ and Fe 3+ were added to the GO suspension and stirred vigorously for 2 min. Next, concentrated NaOH solution (100 g/L) was added into the mixture until the solution pH was 10, then the mixture was stirred constantly for 45 min at 85˚C. The product was rinsed with Milli-Q water to obtain a black-colored GO@Fe 3 O 4 suspension.
Grafting ethylenediamine to the magnetic graphene oxide composite (EDA-GO@Fe 3 O 4 ) was achieved by modifying GO@Fe 3 O 4 with ethylenediamine [24]. First, 9.0 mL ammonia solution was added to the GO@Fe 3 O 4 suspension and stirred for 5 min. Then 36 mL ethylenediamine was added into the suspension and stirred for 10 min. Next, the suspension was stirred at 95˚C for 6 h. Finally, ethanol and Milli-Q water were used to wash the product to neutral pH.

Characteristics of EDA-GO@Fe 3 O 4
High-resolution transmission electron microscopy (HRTEM) of the EDA-GO@Fe 3 O 4 was collected with a Tecnai G2-F20 (FEI, USA). The EDS spectrum was collected with an energy-dispersive X-ray spectrometer (FEI, USA). The FT-IR spectrum of EDA-GO@Fe 3 O 4 was collected using a Magna-IR 170 spectrometer with KBr pellets at room temperature (Nicolet, USA). TG and DSC curves were recorded using a Q600 thermoanalyzer (TA, USA). The surface elemental composition was analyzed using an ESCALAB 250Xi X-ray photoelectron spectroscope with a resolution of 0.5 eV (Thermo, USA).

Adsorption experiments
Batch adsorption experiments. Adsorption experiments were study in a water bath shaker. The EDA-GO@Fe 3 O 4 or GO@Fe 3 O 4 and the Cr(VI) solution were added to 100 mL Erlenmeyer flasks. 0.01 or 0.1 M NaOH and HCl solution was used to adjust the pH values of the suspensions. Then, the Erlenmeyer flasks were shaken for 8 h at the desired temperature. After the adsorption process, a permanent magnet was used to separate the suspension. The concentration of Cr(VI) ions was determined by a UV spectrophotometer at 540 nm [20]. The adsorption capacities (q e , mg/g) and adsorption percentages (E e , %) of EDA-GO@Fe 3 O 4 or GO@Fe 3 O 4 were calculated by the following equations: where C 0 (mg/L) is the initial Cr(VI) concentration; C e (mg/L) is the equilibrium concentration of Cr (
where q t (mg/g) is the adsorption capacity of EDA-GO@Fe 3 O 4 or GO@Fe 3 O 4 at time t (h); k 1 (1/min), k 2 (g/mg min), and k p (mg/g min 0.5 ) are the adsorption rate constants for the three kinetic models, respectively; q e (mg/g) is the adsorption capacities calculated by the kinetics models; C of adsorption constant is the intercept. Adsorption isotherm. The nonlinear form of Langmuir, Freundlich, Temkin isotherm models are given by the Eqs 6, 7 and 8, respectively [27][28][29].
where q e (mg/g) is the adsorption amount of Cr(VI) ions; q max (mg/g) is the maximum adsorption capacities of the adsorbent; K L (L/mg), K F and n, a T (L/g) and b T (kJ/mol) are the constants for the Langmuir, Freundlich, and Temkin isotherm models, respectively; C e (mg/L) is the equilibrium concentration after the adsorption process; T (K) is the temperature; and R (8.314 × 10 −3 kJ/mol K) is the gas constant.

Results and discussion Characterization
The morphology and microstructure of EDA-GO@Fe 3 O 4 were characterized by HRTEM, and images at different magnifications are shown in Fig 1. EDA-GO@Fe 3 O 4 revealed a typical fabric-like shape with a two-dimensional nanosheet structure (Fig 1A and 1B). From Fig 1, several small black spots (Fe 3 O 4 nanoparticles) are dispersed on the GO nanosheets. The EDS spectrum (Fig 2A) indicated that Fe, C, O, and N were present. Carbon came mainly from the GO nanosheets and the oxygen from the oxygen-containing functional groups in GO and Fe 3 O 4 nanoparticles. N arose mainly from the amino groups in the grafted ethylenediamine, which indicated that ethylenediamine had been successfully introduced into the GO@Fe 3 O 4. Fig 2B shows the FT-IR spectrum of EDA-GO@Fe 3 O 4 composite. The peak at 540 cm -1 is attributed to Fe-O in Fe 3 O 4 , indicating successful connections between the Fe 3 O 4 nanoparticles and the GO nanosheets. The peaks at 1571 cm -1 and 1202 cm -1 are attributable to the N-H stretching vibration (in the C-NH group) and C-N (in the-C-NH-C-and-NHCO-groups), respectively [24].
The TG-DSC curves of EDA-GO@Fe 3 O 4 are shown in Fig 2C. The 15% loss of mass between 30˚C and 106˚C was attributed to the evaporation of water. The 51% loss of mass observed when the temperature ranged from 106˚C to 1000˚C was ascribed to pyrolysis of the grafted ethylenediamine and the oxygen-containing functional groups on the surfaces of the EDA-GO@Fe 3 O 4 composite.
In order to gain further information on the chemical composition of EDA-GO@Fe 3 O 4 , XPS analysis was performed; the results are shown in Fig 2D-2F. From the XPS survey scan

Effect of pH
Solution pH is a very important factor for affecting sorption efficiency. In Fig 3, the sorption capacities and adsorption percentages of EDA-GO@Fe 3 O 4 decreased significantly when the solution pH value increased from 2 to 10. For instance, the adsorption capacity and adsorption percentages were 37.73 mg/g and 84% at pH = 2, but only 1.60 mg/g and 4% at pH = 10, respectively. This result indicated clearly that the adsorption process was pH dependent, which may be due to that pH can affect the surface binding sites of EDA-GO@Fe 3   , which may be mainly ascribed to the NO 3 − competing with the HCrO 4 − ions for the adsorption sites [31]. Besides, the NO 3 − might decrease the zeta potentials of the removal [35,36].

Desorption and regeneration analysis
In order to determine the reusability of EDA-GO@Fe 3 O 4 , the adsorption-desorption cycles were conducted for four times, and the results are illustrated in S4 Fig. After   The adsorption experimental data was interpreted by the kinetics models of pseudo-firstorder, pseudo-second-order, and intra-particle diffusion (Fig 4B and 4C). The parameters of pseudo-first-order and pseudo-second-order adsorption kinetics are summarized in Table 1. In both Fig 4B and Table 1, the correlation coefficient (R 2 ) values of the GO@Fe 3 O 4 and EDA-GO@Fe 3 O 4 for the pseudo-second-order model were higher than those for the pseudo-first-order model. In addition, the calculated q e values for the pseudo-second-order model were very close to the experimental data. These results indicate that the pseudo-second-order model is more suitable to describe the adsorption experimental data, indicating that chemical adsorption reaction is the dominant rate-limiting step for both adsorption processes. Fig 4C illustrates the q t vs. t 0.5 plot. The multi-linear plot can be separated into two largely linear regions, indicating that intra-particle diffusion of the Cr(VI) ions was not the rate-controlling step for the overall adsorption process. The first region in Fig 4C (labeled "I") might be assigned to film diffusion corresponding to transportation of Cr(VI) ions from the aqueous solution to the external surfaces of GO@Fe 3 O 4 and EDA-GO@Fe 3 O 4 . The second region (labeled "II") includes the gradual sorption and equilibrium stages [38].

Adsorption isotherm
The decontamination of Cr(VI) ions by EDA-GO@Fe 3 O 4 was studied at 15˚C, 30˚C, and 50˚C to determine the relative parameters of the adsorption isotherms, and the results are demonstrated in Fig 5. The adsorption capacities of EDA-GO@Fe 3 O 4 for Cr(VI) ions increased when the temperature increased from 10 to 50˚C, implying an endothermic process. The nonlinear form of Langmuir, Freundlich, and Temkin adsorption isotherms at different temperatures are also illustrated in Fig 5. The parameters for the three isotherm models are demonstrated in Table 2. The Freundlich model clearly describes the isotherm adsorption data better than the Langmuir and Temkin models within the studied temperature range. The Freundlich constants of n (2.150 for 10˚C, 3.086 for 30˚C, and 3.259 for 50˚C) are within the beneficial adsorption range (1-10) [39], indicating that EDA-GO@-Fe 3 O 4 can be applied as an effective adsorbent. The effects of experimental factor variation and factor interactions on Cr(VI) decontamination were evaluated by significance testing of the FFD model [40]. The half-normal probability   The predicted values are very close to the experimental measurements, implying that the Cr(VI) adsorption process can be predicted by the FFD models obtained. Fig 6C shows that the internally studentized residuals were equally scattered between −3 and +3, which indicated that the obtained FFD model in this study was adequate [40].
The identification of important factors, and factor interaction on Cr(VI) adsorption by EDA-GO@Fe 3 O 4 , is illustrated in Fig 6D. The factors with negative effects were A (−54.00), D (−0.51), and E (−9.85), while positive estimates of the effects of B (25.89) and C (2.52) were observed. This indicates that factor A is very important in the removal process. The effects of the six selected factors on Cr(VI) adsorption by EDA-GO@Fe 3 O 4 were found to lie in the order A > B > E > C > D. Fig 7 shows factor interaction effects for Cr(VI) decontamination. Lines in cells AB, AE, and BC were non-parallel, indicating that these factors could affect each other significantly [35]. In row A, Cr(VI) adsorption decreased slightly with the increase of pH values, indicating