Enhanced removal of roxarsone by Fe₃O₄@3D graphene nanocomposites: synergistic adsorption and mechanism

Abstract

Roxarsone (ROX) is an emerging arsenic pollutant due to its potential degradation into highly toxic inorganic arsenic species in the environment. Adsorbents which can capture ROX with both high capacity and affinity are urgently needed. Herein, a nanocomposite of nano-Fe₃O₄ and three-dimensional graphene (Fe₃O₄@RGO) was designed, aiming to simultaneously attract arsenate and benzene groups in ROX. Characterization of the nanocomposite revealed that Fe₃O₄ nanoparticles (20–50 nm) with exposed (400) planes were highly dispersed on the graphene support. Adsorption experiments showed that Fe₃O₄@RGO had higher adsorption capacity, affinity, and adsorption rate towards ROX than pristine materials and efficiently removed ROX from both simulated natural and waste waters. The adsorption mechanism was confirmed as a synergetic interaction of As–Fe coordination, hydrogen bonding and π–π interaction. X-ray photoelectron spectroscopy (XPS) and extended X-ray absorption fine structure (EXAFS) results suggested that the As–Fe complexes and hydrogen bonds between Fe₃O₄@RGO and ROX were stronger than those in pristine nano-Fe₃O₄, due to the greater number of surface hydroxyls and shorter As–Fe atomic distance in Fe₃O₄@RGO. The π–π interaction between ROX and the graphene part in Fe₃O₄@RGO was also enhanced. This study provided a novel idea for designing materials to remove pollutants with both inorganic and organic moieties, such as phenylarsonic acid compounds, from water.

---

Environmental significance

Roxarsone (ROX) is considered to be an emerging pollutant as it can degrade into highly toxic inorganic arsenic species in natural environments. Reported in this work is the design of a Fe₃O₄@RGO nanocomposite to simultaneously adsorb the arsenate and benzene ring groups in ROX molecules. The excellent performance of Fe₃O₄@RGO achieved the expected high adsorption capacity and affinity for ROX, as well as the superior ROX removal in both simulated natural and waste waters, with no ROX degradation. To our knowledge, this is the first molecular-level study to investigate the synergistic adsorption between the designed adsorbent and the two functional groups in organoarsenic compounds. The enhanced removal of ROX was ascribed to the synergetic interactions of As–Fe coordination, hydrogen bonding and π–π stacking between Fe₃O₄@RGO and ROX. The combination of 3D graphene and Fe₃O₄ nanoparticles increased the surface hydroxyl amounts in the nanocomposite, and thus further enhanced the As–Fe coordination and hydrogen bonding for superior adsorption. On this basis, this work is expected to stimulate further investigations into the functional design of adsorbents for the removal of emerging pollutants with both inorganic and organic moieties.

1 Introduction

Roxarsone (ROX) is a typical phenylarsonic acid compound which is widely used as an animal feed additive due to its broad antimicrobial properties.^1,2 After excretion in manure, water-soluble ROX can be degraded into more toxic inorganic arsenic species (such as arsenate) in 30 days via biotic and abiotic degradation, leading to severe arsenic pollution in soil and groundwater.^3–5 It was reported in China that the arsenic content in chicken litter and pig manure was 21.6 and 89.3 mg kg⁻¹, respectively,⁶ which came from ROX and other arsenic-containing feed additives. Therefore, ROX is considered as a type of emerging contaminant, and its removal is crucial for maintaining water and soil quality and safety.

Since the degradation product of ROX is highly toxic, adsorption is preferable over degradation. In particular, adsorbents with both high capacity and affinity for ROX are urgently needed to restrain the release of ROX during and after adsorption. Investigations so far have mainly focused on materials with high adsorption capacity for ROX.^7–9 For example, metal–organic frameworks (MOFs) were found with high adsorption capacity for ROX (up to 730 mg g⁻¹).¹⁰ However, the adsorption affinity of MOFs for ROX was relatively low, although the adsorption mechanism included the coordination between the open metal sites and arsenic groups, electrostatic interaction and pore filling effects.¹¹ Moreover, very few studies assessed whether ROX degraded into inorganic arsenic species during adsorption. Therefore, new materials with both high adsorption capacity and strong affinity for ROX need to be designed, synthesized and tested.

Fe₃O₄ nanoparticles (NPs), which were confirmed to be excellent adsorbents for arsenic,¹² could be a potential candidate for removing ROX because the ROX molecule contains an As-bearing group similar to arsenate (Fig. S1 in the ESI†). According to previous studies,^13–15 the high affinity of Fe₃O₄ for arsenate was attributed to the inner-sphere coordination of bidentate binuclear corner-sharing (²C) or tridentate hexanuclear corner-sharing (³C) complexes between Fe and As. Moreover, the {100} facet (i.e., (400) plane) of Fe₃O₄ was considered to be favorable for the formation of As–Fe ²C complexes.¹⁶ It was also suggested that hydroxyls on the surface of Fe₃O₄ facilitated the inner-sphere coordination between Fe and As,¹⁷ thereby greatly enhancing the adsorption affinity. However, the spontaneous agglomeration of nano-Fe₃O₄ resulted in poor dispersibility in water, causing a significant decrease in surface active sites for binding As, thus severely limiting its application.¹⁸

New adsorbents with high affinity for ROX could be designed based on the features of arsenate and benzene ring groups in the ROX molecule (Fig. S1†). The nanocomposite of Fe₃O₄ NPs loaded on three-dimensional (3D) reduced graphene oxide (Fe₃O₄@RGO) was selected on account of the following hypotheses. Firstly, 3D graphene with an interior connected porous structure has been confirmed to be an excellent support for nano-Fe₃O₄ to refrain the NPs from aggregation.^19–21 By immobilization on 3D graphene, more active sites on Fe₃O₄ NPs could become accessible. Secondly, the graphene substrate may promote the nucleation and inerratic growth of Fe₃O₄ NPs, which not only decreases the grain size but also possibly changes the preferential crystallographic facet and increases the amount of hydroxyl groups on the surface. It is expected that the preferential facet and high amount of surface hydroxyls in the nanocomposite would enhance the As–Fe coordination. Thirdly, graphene in the nanocomposite is also expected to adsorb ROX through π–π stacking with the benzene ring groups.²² Once ROX is adsorbed by the As–Fe short-distance interaction and/or other forces, it is very possible that the graphene support in the nanocomposite would undergo π–π interaction with the benzene ring of the same ROX molecule, which would further enhance the interaction between the adsorbent and ROX. Thus, Fe₃O₄ NP–graphene composites may provide potential synergy for ROX adsorption. Currently, although a few iron-based carbon materials have been used for ROX adsorption,^8,23 without a rationally justified design, the merit of these materials could not be well manifested. Then, the roles of iron and carbon in ROX adsorption could be only discussed separately, and no synergetic adsorption mechanisms were considered in these works.

In this work, a Fe₃O₄@RGO nanocomposite was designed and prepared via in situ crystal growth of Fe₃O₄ NPs on RGO. The main objectives of this work were therefore to (1) analyze the dispersibility and preferential exposed facets of Fe₃O₄ NPs on RGO, as well as the chemical bonds between Fe₃O₄ and RGO, (2) comparatively investigate the adsorption of ROX on Fe₃O₄@RGO and the original pristine materials to verify the synergetic adsorption, and evaluate the ROX removal potential of Fe₃O₄@RGO from simulated natural and waste waters, and (3) disclose the factors, such as exposed planes and surface hydroxyls of Fe₃O₄ in the nanocomposite that are responsible for the formation of As–Fe complexes and hydrogen bonds, and assess the contribution of As–Fe coordination, hydrogen bonding, π–π stacking and electrostatic interaction to the overall adsorption. This study will provide an important theoretical and experimental basis for designing adsorbents with high adsorption capacity and affinity towards ROX.

2 Materials and methods

2.1 Synthesis and characterization of the adsorbents

Fe₃O₄@RGO was synthesized using an in situ compositing method, via a redox process between ferrous ions and graphene oxide (GO).²¹ Briefly, a 2 mg mL⁻¹ GO suspension was mixed with 1.0 mmol FeSO₄·7H₂O at pH 10 and then kept at 90 °C for 6 h to produce a self-assembled nanocomposite (details are described in ESI† Section S1). For comparison, pure Fe₃O₄ NPs synthesized using the same method without GO were named p-Fe₃O₄, while GO treated using the same procedure without the addition of Fe²⁺ was named GO-90. Preliminary results revealed that the GO could be only slightly reduced at 90 °C, while treatment under hydrothermal conditions at an elevated temperature of 180 °C facilitated the reduction of GO at the same level as that in Fe₃O₄@RGO, thus it was named RGO (Fig. S2 and Table S1†).

The crystal structure, surface morphology, specific surface area and surface charge of the samples were analyzed using X-ray diffraction (XRD), scanning electron microscopy (SEM), the Brunauer–Emmett–Teller (BET) N₂ adsorption method and zeta potential analysis, respectively. The active crystal planes of Fe₃O₄ in the nanocomposite and p-Fe₃O₄ were examined by high-resolution transmission electron microscopy (HRTEM). Surface elemental composition and functional groups were determined using X-ray photoelectron spectroscopy (XPS) and Raman spectroscopy. The mass content of Fe₃O₄ in Fe₃O₄@RGO was determined using a modified ferrozine method.²⁴ Details about the characterization methods are presented in ESI† Section S1.

2.2 Adsorption experiments

Adsorption isotherm studies were conducted with ROX (98%, Aladdin Bio-Chem Technology, China) concentrations from 1 to 100 mg L⁻¹, while the adsorbent concentrations of Fe₃O₄@RGO, p-Fe₃O₄, GO-90 and RGO were kept at 0.15 g L⁻¹ in 20 mL solution. The mixtures were shaken at pH 5 ± 0.1 and 25 °C with a speed of 200 rpm for 72 h. After magnetic separation,²⁵ the residual ROX concentrations were measured using a UV-vis spectrometer (UV-2600, Shimadzu) at 223 nm with a solution pH of 4.0 ± 0.1. The possible degradation of ROX during adsorption was investigated by determining the concentrations of inorganic As(V) and As(III) in the aqueous solution using HPLC-ICP-MS (ESI† Section S1).²⁶ The adsorption isotherms were fitted with Dubinin–Astakhov (DA), Langmuir and Freundlich models.^27,28 For kinetic experiments, 50 mg L⁻¹ ROX solution was mixed with 0.15 g L⁻¹ adsorbents in 300 mL solution and shaken at pH 5 ± 0.1 and 25 °C with a speed of 200 rpm for 1 min to 4 days. The adsorption kinetics was fitted with a pseudo-second-order non-linear kinetic model. The effects of pH on adsorption were investigated by conducting adsorption experiments within the pH range of 3.0–11.0, while the pH values were adjusted with NaOH and HCl solutions. To further investigate the potential applications of Fe₃O₄@RGO in natural and waste waters, ROX adsorption in the presence of dissolved organic matter (DOM, 0–12 mg C L⁻¹) and in swine manure lixivium (with a DOM concentration of 9.0 mg C L⁻¹ and an added ROX concentration of 1 mg L⁻¹) was studied. Experimental and analytical details are displayed in Section S1.† The desorption and reusability of Fe₃O₄@RGO were determined after washing the used adsorbent with 0.05 mol L⁻¹ NaOH for 1 h at 25 °C with shaking at a speed of 200 rpm.²⁹

2.3 Adsorption mechanism analysis

ROX adsorption as a function of solution pH (3–12) was conducted to estimate the contribution of electrostatic interaction between ROX and the nanocomposite. An ROX solution with a concentration of 50 mg L⁻¹ was mixed with 0.15 g L⁻¹ adsorbents and shaken at 25 °C with a speed of 200 rpm for 72 h. The As–Fe surface complex was analyzed by extended X-ray absorption fine structure spectroscopy (EXAFS). Briefly, the As K-edge X-ray absorption fine structure (XAFS) spectra of As-loaded samples were acquired at beamline 4W1B of the Beijing Synchrotron Radiation Facility (BSRF), China. The hydrogen bonding between the adsorbents and ROX was analyzed using Fourier transform infrared spectroscopy (FTIR). The hydroxyl content in Fe₃O₄@RGO and p-Fe₃O₄ was calculated from the O 1s XPS spectra, while the π–π stacking of Fe₃O₄@RGO and RGO after ROX adsorption was determined from the C 1s spectra. Detailed analysis procedures are provided in Section S1.†

3 Results and discussion

3.1 Characterization of the Fe₃O₄@RGO nanocomposite

The XRD patterns of Fe₃O₄@RGO exhibited the characteristic peak of graphite (2θ = 13.5°) and the characteristic peaks of Fe₃O₄ (ICSD No. 84611, Fig. S3†). XPS analysis of the Fe 3p spectra also confirmed the presence of Fe₃O₄ in the nanocomposite (Fig. S4†). These results suggested the excellent compositing of Fe₃O₄ NPs with graphene. The mean crystal size of Fe₃O₄ NPs in the nanocomposite calculated from the Scherrer equation was around 25 nm, which was obviously smaller than that of p-Fe₃O₄ (around 34 nm, Table S2†), indicating that the in situ growth on the graphene substrate could effectively decrease the crystal size of Fe₃O₄ NPs. The weight ratio of Fe₃O₄ and graphene in the nanocomposite was around 4.05 : 1. The specific surface area of Fe₃O₄@RGO, p-Fe₃O₄ and GO-90 was 61.9, 103.1 and 28.1 m² g⁻¹, respectively (Table S2†).

SEM observation revealed that GO-90 had a complete sheet-like structure with a wrinkled surface (Fig. 1a), while RGO had more pores in the sheets (Fig. 1b), indicating that the increased temperature for achieving a high reduction degree could be at the expense of losing morphology features. p-Fe₃O₄ exhibited irregular and stacked nanoplates with lateral sizes of 200–400 nm (Fig. 1c). Fe₃O₄@RGO showed a similar sheet-like structure to GO-90, while Fe₃O₄ NPs with diameters of 30–50 nm were uniformly immobilized on the graphene sheets (Fig. 1d). These results indicated that the graphene support was beneficial for decreasing the particle size of Fe₃O₄ NPs and restricting their aggregation. In addition, HRTEM images showed that the d-spacing of lattice fringes for p-Fe₃O₄ and Fe₃O₄@RGO was 0.48 and 0.21 nm, respectively (Fig. 1e and f), suggesting that the (111) plane is mainly exposed in p-Fe₃O₄ while the (400) plane is mainly exposed in the Fe₃O₄ NPs of the nanocomposite. Further studies via the O 1s XPS spectra and Raman spectra confirmed that the immobilization of Fe₃O₄ NPs on graphene had a significant feature of Fe–O–C bonding (Fig. S5†), which agreed with previous studies,^30,31 indicating the strong interaction between Fe₃O₄ and the graphene support.

Fig. 1 SEM images of GO-90 (a), RGO (b), p-Fe₃O₄ (c), and Fe₃O₄@RGO (d) and HRTEM images of p-Fe₃O₄ (e) and Fe₃O₄@RGO (f).

3.2 Adsorption behavior of Fe₃O₄@RGO towards ROX

3.2.1 Adsorption isotherms. Adsorption of ROX on Fe₃O₄@RGO was investigated, along with p-Fe₃O₄, GO-90 and RGO for comparison. The adsorption isotherms of ROX on the tested materials are presented in Fig. 2a. Obviously, Fe₃O₄@RGO displayed higher adsorption capacity (q_e) than the three pristine materials. After normalizing with surface area, the adsorption capacity of Fe₃O₄@RGO was further elevated (Fig. 2b), indicating that the enhanced ROX adsorption on Fe₃O₄@RGO was independent of the surface area. Better fitting was observed using DA and Langmuir models than that using the Freundlich model (Table 1). As indicated by DA fitting, the maximum adsorption capacity (Q_m) followed the order of Fe₃O₄@RGO > p-Fe₃O₄ > p-Fe₃O₄ + GO-90 (calculated from the physical mixture of p-Fe₃O₄ and GO-90 with a mass ratio of 4.05 : 1) > RGO > GO-90, while the adsorption affinity (as indicated by E values) followed the order of Fe₃O₄@RGO ≈ p-Fe₃O₄ > p-Fe₃O₄ + GO-90 > RGO > GO-90. Therefore, the combination of Fe₃O₄ NPs and graphene in Fe₃O₄@RGO displayed a synergetic enhancement on both adsorption capacity and affinity for ROX.

Fig. 2 Adsorption isotherms (a), surface area-normalized adsorption isotherms (b), and adsorption kinetics (fitted by the pseudo-second-order non-linear kinetic model) (c) of ROX on Fe₃O₄@RGO, p-Fe₃O₄, GO-90 and RGO. In panel (a), the magenta symbols () stand for the calculated adsorption capacity value of the physical mixture of p-Fe₃O₄ and GO-90.

Table 1 Parameters of Langmuir, Freundlich, and DA fits for ROX adsorption on Fe₃O₄@RGO, p-Fe₃O₄ GO-90, and RGO

Sample	Langmuir model			Freundlich model			DA model
	Q m (mg g^−1)	K L	R ^2	n	K f (mg g^−1)	R ^2	Q m (mg g^−1)	E	b	R ^2
a The Langmuir, Freundlich, and DA fitting of p-Fe3O4 + GO-90 was conducted from the additive results of residual concentration and adsorption capacity obtained from 80.2% p-Fe3O4 and 19.8% GO-90.
Fe3O4@RGO	454.48	0.78	0.9975	2.70	134.80	0.9562	534.43	4.72	2.44	0.9939
p-Fe3O4	163.93	0.75	0.9930	8.56	69.96	0.9212	172.16	4.81	2.49	0.9999
GO-90	43.10	0.025	0.9948	1.51	6.73	0.9915	60.02	2.42	1.69	0.9980
p-Fe3O4 + GO-90^a	111.11	0.044	0.9936	2.05	24.05	0.8911	136.48	3.87	3.27	0.9941
RGO	104.17	0.065	0.9879	1.70	9.71	0.9738	92.78	3.42	2.36	0.9986

---

The comparison of the Langmuir-fitted maximum adsorption capacity (Q_m) and affinity (K_L) of Fe₃O₄@RGO with other adsorbents in the literature is shown in Table S3.† It can be seen that although some of the adsorbents had larger Q_m, most of the K_L values were smaller than those for Fe₃O₄@RGO, and thus Fe₃O₄@RGO had a higher adsorption amount at equilibrium in both low and high concentration systems. Moreover, except the ferric and manganese binary oxide,³² none of the reported adsorbents considered the degradation of ROX during adsorption. In comparison, for the as-prepared Fe₃O₄@RGO, no inorganic As(V) or As(III) was detected in the ROX adsorption supernatant even after 15 days or in the desorption solution in water and NaOH (Fig. S6†). Therefore, compared with the reported adsorbents, Fe₃O₄@RGO was a more effective and environmentally friendly adsorbent towards ROX with both high adsorption capacity and affinity, as well as no risk of releasing inorganic arsenic into the environment.

3.2.2 Adsorption kinetics. The adsorption kinetics of ROX on the tested materials are shown in Fig. 2c. Adsorption equilibrium was reached around 720 min for Fe₃O₄@RGO, whereas longer times were required for the other samples. All the adsorption kinetics followed the pseudo-second-order model (Table S4†), indicating that chemical sorption controlled the sorption rate and the number of active sites on the sorbent determined the adsorption capacity.³³ The rate constant (k₂) of Fe₃O₄@RGO was 2.38 × 10⁻⁴ g mg⁻¹ min⁻¹, which was higher than those of the three pristine materials (Table S4†). The enhanced adsorption rate of Fe₃O₄@RGO, together with its higher adsorption capacity and affinity, was possible due to the stronger chemical interaction between Fe₃O₄@RGO and ROX, which is discussed in detail in the Adsorption mechanism of Fe₃O₄@RGO towards ROX section.

3.2.3 Reusability of Fe₃O₄@RGO. In this work, NaOH solution was chosen as the desorption solution to test the reusability of Fe₃O₄@RGO. The result indicated that the adsorption capacity of the used Fe₃O₄@RGO did not diminish significantly after four cycles (Fig. S7†). FTIR analysis clearly showed the same characteristic peaks of Fe₃O₄@RGO after the fourth adsorption cycle with the pristine sample, while XRD patterns also showed the characteristic peaks of Fe₃O₄ after the fourth cycle (Fig. S8†). This result suggested the stable structure properties of Fe₃O₄@RGO during the repeated regeneration process.

3.2.4 Performance of Fe₃O₄@RGO in simulated natural and waste waters. DOM (dissolved organic matter) is the major inhibitor in natural water for the adsorption of organic pollutants.³⁴ Therefore, the effect of DOM on ROX adsorption was investigated and the results are shown in Fig. 3a. The ROX removal rate slightly decreased from 98.6% to 93.5% when the DOM concentration increased from 0 to 12 mg C L⁻¹. Since DOM is present at levels of milligrams per liter in natural water, the slight decrease in ROX removal rates suggested that Fe₃O₄@RGO could be potentially applied to ROX removal in natural water.

Fig. 3 (a) Effect of DOM on the removal rate of ROX adsorbed on Fe₃O₄@RGO, with an initial ROX concentration of 1 mg L⁻¹. (b) The removal rate of ROX and the residual arsenic concentration in a natural swine manure lixivium with Fe₃O₄@RGO dosages of 0.15 and 0.30 g L⁻¹.

Since swine manure is a significant source of ROX in wastewater and the composition of organic matter in swine manure could represent certain real wastewater,³⁵ the removal of ROX in swine manure lixivium was investigated to evaluate the potential application of Fe₃O₄@RGO in real wastewater. As shown in Fig. 3b, the 90 min removal rate of ROX by Fe₃O₄@RGO at dosages of 0.15 and 0.30 g L⁻¹ reached 92% and 99%, respectively. Moreover, the dosage of 0.30 g L⁻¹ Fe₃O₄@RGO decreased the arsenic concentration in ROX from 285 to 3.1 μg L⁻¹, which satisfied the arsenic standard in surface water of China (50 μg L⁻¹, GB3838-2002), suggesting the excellent removal of ROX by Fe₃O₄@RGO in waste water. Therefore, Fe₃O₄@RGO is potentially applied to ROX removal in both natural and waste waters.

3.3 Adsorption mechanism of Fe₃O₄@RGO towards ROX

3.3.1 Identification of involved adsorption mechanisms.
Electrostatic interaction. To reveal the role of electrostatic interaction during adsorption, the correlation between ROX adsorption capability and pH values for all the four adsorbents was conducted. As shown in Fig. S9,† Fe₃O₄@RGO presented the highest adsorption capacity at pH 5.0, which then sharply decreased with increasing pH. p-Fe₃O₄ displayed a similar trend to Fe₃O₄@RGO, although the maximum adsorption occurred at pH 7. Zeta potential measurements showed that the point of zero charge (pH_PZC) for Fe₃O₄@RGO, p-Fe₃O₄, RGO and GO-90 was at pH 5.8, 7.3, 2.9 and <2, respectively (insert of Fig. S9†). Thus, the maximum adsorption of both Fe₃O₄@RGO and p-Fe₃O₄ occurred almost at the pH_PZC, indicating that electrostatic attraction was not the primary mechanism for ROX adsorption on Fe₃O₄@RGO at pH 5. In addition, ROX was mainly negatively charged in the tested pH range (3–11) since the aqueous dissociation constants of ROX molecules were 3.49, 5.74 and 9.13, respectively (Fig. S1†). Due to similar values between the pK_a (5.74) of ROX and the pH_PZC of Fe₃O₄@RGO, negative charge-assisted hydrogen bonds may exist between Fe₃O₄@RGO and ROX, which was reported between carboxylated and hydroxylated carbon nanotubes and weak acids (e.g., phthalic acid).³⁶ Meanwhile, negatively charged GO-90 and RGO displayed decreasing adsorption capacity in the tested pH range, but still showed a certain level of ROX adsorption. Therefore, other interactions in addition to electrostatic interaction must exist between the graphene support and ROX molecules.
π–π interaction. The C 1s XPS spectra were analyzed to reveal the interaction between the graphene support and ROX in Fe₃O₄@RGO, setting RGO as a comparison (Fig. 4). The C–C/C=C band proportion of Fe₃O₄@RGO was close to that of RGO, indicating a similar reduction level of GO in the two materials. After adsorbing ROX, the band proportion increased from 74.61% to 90.52% and 76.41% to 82.77% for Fe₃O₄@RGO and RGO, respectively. The increase in this band proportion was attributed to the π–π interaction between the graphene support and ROX molecules. Moreover, the increase range of the band proportion in Fe₃O₄@RGO was higher than that in RGO, suggesting a stronger π–π interaction between Fe₃O₄@RGO and ROX, which might result from the synergistic effect of electrostatic attraction. At pH 5, Fe₃O₄@RGO was still positively charged while RGO was negatively charged. Such a discrepancy resulted in the difference of the electrostatic interactions between ROX and the two adsorbents. In other words, the weak electrostatic attraction would narrow the molecule distance between Fe₃O₄@RGO and ROX, thus promoting the formation of π–π stacking between the graphene part and ROX.

Fig. 4 C 1s spectra of Fe₃O₄@RGO (a) and RGO (b) before and after ROX adsorption.

As–Fe complex. In view of the higher adsorption capacity of p-Fe₃O₄ (163.93 mg g⁻¹) than that of GO-90 (43.10 mg g⁻¹) and the dominant content of Fe₃O₄ (80.2%, w/w) in the nanocomposite, the adsorption of ROX on Fe₃O₄@RGO was considered mainly from the Fe₃O₄ NPs in the nanocomposite. The As–Fe interaction between Fe₃O₄@RGO and ROX was studied via XAFS analysis. The X-ray absorption near-edge structure (XANES) data together with the As 3d spectra from the XPS results confirmed that no ROX was degraded into inorganic As(V) or As(III) during adsorption by Fe₃O₄@RGO and p-Fe₃O₄ (Fig. S10†), which was consistent with the HPLC-ICP-MS results. The unfiltered k³-weighted arsenic K-edge EXAFS data and their Fourier transform (FT) spectra can be found in Fig. 5. As shown in Fig. 5b, the R-space spectra consisted of a predominant signal derived from As–O first-neighbor contributions and a weaker signal assigned to As–Fe second-neighbor contributions. The As–O first-neighbor contributions were fitted with 3.2–3.4 oxygen atoms at 1.69 ± 0.02 Å for both samples (Table 2), which agreed with the calculation result of the tetrahedral structure around As in phenylarsonic acid compounds.³⁷ The second-neighbor fitting results indicated that the As atom was surrounded by Fe atoms at 3.30 and 3.35 Å for Fe₃O₄@RGO–ROX and p-Fe₃O₄–ROX (Table 2), respectively, suggesting the formation of ²C complexes in both samples.¹³

Fig. 5 Arsenic K-edge XAFS data of Fe₃O₄@RGO–ROX and p-Fe₃O₄–ROX samples: unfiltered k³-weighted K-space EXAFS data (a). Magnitude part of the FT R-space (b). Real part of the FT R-space (c). Experimental and calculated curves are displayed as red open circles and black solid lines, respectively. Results of the k³-weighted fit are reported with their corresponding FT (Table 2).

Table 2 Results of shell-by-shell fitting of EXAFS data for samples with adsorbed ROX

Sample	Atomic path	CN^a	R (Å)	σ ^2 (Å)	ΔE^d (eV)	R-Factor^e
a CN: number of neighbors. b R (Å): interatomic distances. c σ ^2 (Å): Debye–Waller factor. d ΔE (eV): difference between the user-defined threshold energy and the experimentally determined threshold energy, in electron volts. e R-Factor: goodness of fitting. f As–Fe: As–Fe second-neighbor pairs. g MSAs–O–O: multiple scattering. The second-neighbor contributions of both samples were fitted using As–Fe pairs and an additional multiple-scattering (MS) contribution of As–O–O–As corresponding to Fe(AsO4)·2H2O. The amplitude reduction factor (SO2) was fixed at 0.95. Values in the parentheses are standard errors of the fitted parameters.
Fe3O4@RGO–ROX	As–O	3.2 (0.1)	1.69 (0.02)	0.002 (0.001)	6.4 (1.5)	0.023
	As–Fe^f	1.7 (0.3)	3.30 (0.05)	0.01
	MSAs–O–O^g	12	3.08	0.001
p-Fe3O4–ROX	As–O	3.4 (0.1)	1.69 (0.02)	0.003 (0.001)	5.6 (1.2)	0.021
	As–Fe^f	1.3 (0.2)	3.35 (0.05)	0.008
	MSAs–O–O^g	12	3.10 (0.05)	0.001

---

However, the coordination numbers (CN) of Fe atoms around the As atom in the second-neighbor contribution of Fe₃O₄@RGO–ROX and p-Fe₃O₄–ROX were different (Table 2). Fe₃O₄@RGO–ROX had a higher As–Fe coordination number of 1.7 than p-Fe₃O₄–ROX (1.3). It was reported that the lower number of neighbors could be attributed to the occurrence of outer-sphere complexes.¹⁵ Thus, the higher number of neighbors in Fe₃O₄@RGO–ROX suggested the lower impact of outer-sphere complexes. Meanwhile, the As–Fe distance of Fe₃O₄@RGO–ROX (3.30 Å) was a little shorter than that of p-Fe₃O₄–ROX (3.35 Å), indicating that compositing with graphene narrowed the As–Fe atomic distance during adsorption. Thus, EXAFS analysis confirmed that the compositing of Fe₃O₄ and 3D graphene was beneficial to the formation of an As–Fe inner-sphere complex, and enhanced the bond strength between Fe and As atoms in the complex.

Hydrogen bonds. Hydrogen bonds between the surface hydroxyl groups on the Fe₃O₄ surface (Fe–OH) and ROX molecules also played an important role during the adsorption, since the surface functional groups in Fe₃O₄@RGO were almost electroneutral and the hydroxyl groups in ROX were unionized at pH 5 (Fig. S1†).³⁸ The FTIR spectra of Fe₃O₄@RGO and p-Fe₃O₄ before and after adsorbing ROX are shown in Fig. S11.† The characteristic peaks around 3500 and 1167 cm⁻¹ were assigned to the stretching vibration and bending vibration of Fe–OH, respectively.^39,40 Compared with p-Fe₃O₄, the obvious increase in intensity of the two bands in Fe₃O₄@RGO suggested that more Fe–OH groups existed in the nanocomposite, which was consistent with the XPS results discussed in the following section. It was reported that the formation of hydrogen bonds would cause a red-shift of the –OH vibration peaks.⁴¹ Thus, the red-shift of the Fe–OH characteristic peaks in Fe₃O₄@RGO from 3507 and 1167 cm⁻¹ to 3520 and 1178 cm⁻¹ after ROX adsorption indicated that Fe–OH groups participated in the adsorption as hydrogen bonding sites.

3.3.2 Synergetic adsorption mechanism of Fe₃O₄@RGO. The higher adsorption capacity and affinity of Fe₃O₄@RGO relative to the pristine materials could probably be attributed to the enhanced As–Fe coordination and hydrogen bonding between Fe₃O₄@RGO and ROX molecules. The enhancement of As–Fe coordination may be partly promoted by the exposed facets of Fe₃O₄ NPs in the nanocomposite. It was reported that the dominant ²C complexes were formed between the adsorbed As(V) and the {100} facet of magnetite.¹⁶ As mentioned above, the HRTEM results showed the (400) plane as the major exposed facet of Fe₃O₄ NPs in Fe₃O₄@RGO, which belonged to the {100} facet of magnetite. Therefore, the enhancement of As–Fe coordination might partly result from the (400) planes exposed on the Fe₃O₄@RGO surface. Besides the exposed planes, Fe–OH groups also played a very important role in As–Fe coordination, since previous work revealed that the formation of the As–Fe inner-sphere complex was realized through the replacement of hydroxyl groups in Fe–OH with arsenate.^42,43 As shown in Fig. 6a, the O 1s XPS spectra of Fe₃O₄@RGO before and after adsorption could be deconvoluted into four peaks at 530.5, 531.6, 532.2, and 533.3 eV, which referred to Fe–O/As–O from Fe₃O₄ and ROX, –OH/Fe–O–C, H₂O, and C–O–C, respectively.^44,45 The proportion of –OH peaks in the O 1s spectra was 46.06% and 20.16% for Fe₃O₄@RGO and p-Fe₃O₄, respectively (Fig. 6). Considering the formation of the ²C complex between As and Fe, it was calculated that the Fe–OH species participating in the formation of the As–Fe complex constituted 26.30% of the total oxygen in Fe₃O₄@RGO, which was much higher than that in p-Fe₃O₄ (15.96%, Fig. S12 and Table S5†). This result suggested that more As–Fe complexes could be formed in Fe₃O₄@RGO. Since the O 1s spectra revealed that the proportion of –OH/Fe–O–C groups in the nanocomposite was 46.06%, it could be concluded that the Fe–OH species replaced by arsenate accounted for more than one half of the –OH/Fe–O–C groups in the nanocomposite.

Fig. 6 O 1s XPS spectra of Fe₃O₄@RGO (a) and p-Fe₃O₄ (b) before and after ROX adsorption.

The variation of hydroxyl groups in the nanocomposite also indicated that the hydrogen bonds between Fe₃O₄@RGO and ROX were stronger than those of the pristine p-Fe₃O₄. Deducting the Fe–OH groups forming the As–Fe complex, the proportion of residual hydroxyl groups in p-Fe₃O₄ was only 4.20%, suggesting that the hydrogen bonding between p-Fe₃O₄ and ROX was extremely weak. Meanwhile, the hydroxyl amount in Fe₃O₄@RGO that could form hydrogen bonds with ROX was calculated to be 19.76%. Hence, the higher content of hydroxyl groups in Fe₃O₄@RGO could enhance both the As–Fe coordination and hydrogen bonding between the adsorbent and ROX molecules.

Since the hydroxyl proportion of hydrogen bonds (19.76%) in the total oxygen amount of Fe₃O₄@RGO was lower than that of the As–Fe complex (26.30%, Table S5†), it could be inferred that As–Fe coordination had a higher contribution to ROX adsorption than hydrogen bonding. Moreover, it was reported that the binding energy of the As–Fe complex was around 20 kcal mol⁻¹, which was higher than those of hydrogen bonding (∼8 kcal mol⁻¹) and π–π interaction (∼10 kcal mol⁻¹).^37,46,47 Therefore, considering the different binding energies of the three interactions, as well as the high content of Fe₃O₄ in Fe₃O₄@RGO and high adsorption capacity of Fe₃O₄ for ROX, it could be concluded that the contribution of the three forces to ROX adsorption was in the order: As–Fe coordination > hydrogen bonding > π–π interaction.

A synergistic mechanism for adsorption of ROX on the Fe₃O₄@RGO nanocomposite is illustrated in Fig. 7. As–Fe coordination, hydrogen bonding and π–π stacking were considered to be the dominant forces for ROX adsorption. The co-existence of the graphene support and Fe₃O₄ NPs helped Fe₃O₄ expose (400) planes and shortened the As–Fe atomic distance, which likely promoted the formation of the As–Fe complex, while the graphene support also contributed to the adsorption through π–π stacking. Moreover, the abundant hydroxyl groups present on the surface of Fe₃O₄@RGO enhanced both the As–Fe coordination and hydrogen bonding. The weak electrostatic attraction force brought more ROX molecules close to the substrate and enhanced the π–π stacking. Hence, the synergy between Fe₃O₄ NPs and the graphene support in Fe₃O₄@RGO greatly strengthened the adsorption capability and affinity of Fe₃O₄@RGO for ROX.

Fig. 7 Schematic diagram for the proposed synergetic mechanism of ROX adsorption on Fe₃O₄@RGO. The maximum adsorption occurred at pH 5, at which the surface of Fe₃O₄@RGO was almost electroneutral. The dominant adsorption forces were considered to be As–Fe coordination, hydrogen bonding and π–π stacking.

4 Conclusions

In this work, a Fe₃O₄@RGO nanocomposite was designed for effective removal of ROX which is an emerging arsenic pollutant. To our knowledge, this is the first molecular-level study to investigate the synergistic adsorption between the designed adsorbent and the two functional groups of phenylarsonic acid compounds. The excellent performance of the Fe₃O₄@RGO nanocomposite achieved higher adsorption capacity and affinity for ROX, compared with pristine p-Fe₃O₄, RGO and GO-90, and no ROX was degraded into inorganic arsenic after adsorption and desorption. Moreover, the ROX adsorption rates of Fe₃O₄@RGO in DOM-containing solution and swine manure lixivium were both over 90%, suggesting Fe₃O₄@RGO as a potential adsorbent to remove ROX in both natural and waste waters. The enhanced removal of ROX was ascribed to the synergistic As–Fe coordination, hydrogen bonding and π–π stacking between the adsorbent and adsorbate. In addition, graphene in the nanocomposite not only promoted the dispersibility of Fe₃O₄ NPs, but also increased the surface hydroxyl amounts, thus enhancing the As–Fe coordination and hydrogen bonding for superior adsorption. On this basis, this work is expected to stimulate further investigations into the functional design of adsorbents for the removal of emerging pollutants with both inorganic and organic moieties.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the China Postdoctoral Science Foundation (no. 2016M600654), the National Natural Science Foundation of China (grant no. 21477129), the Guangdong Innovative and Entrepreneurial Research Team Program (No. 2016ZT06N569), and the Fundamental Research Funds for the Central Universities (no. 2017PY009 and 2017BQ054). The authors thank the beamline 4W1B (Beijing Synchrotron Radiation Facility) for providing the beam time.

References

1. J. R. Garbarino, A. J. Bednar, D. W. Rutherford, R. S. Beyer and R. L. Wershaw, Environ. Sci. Technol., 2003, 37, 1509–1514.
2. F. T. Jones, Poult. Sci., 2007, 86, 2–14.
3. L. Wang and H. Cheng, Environ. Sci. Technol., 2015, 49, 3473–3481.
4. E. D'Angelo, G. Zeigler, E. G. Beck, J. Grove and F. Sikora, Sci. Total Environ., 2012, 438, 286–292.
5. J. Han, F. Zhang, L. Cheng, Y. Mu, D. Liu, W. Li and H. Yu, Environ. Sci. Technol. Lett., 2017, 4, 350–355.
6. L. Yao, G. Li and Z. Dang, Yingyong Shengtai Xuebao, 2006, 17, 1989–1992.
7. X. Zhu, F. Qian, Y. Liu, D. Matera, G. Wu, S. Zhang and J. Chen, Carbon, 2016, 99, 338–347.
8. X. Zhu, F. Qian, Y. Liu, S. Zhang and J. Chen, RSC Adv., 2015, 5, 60713–60722.
9. J. Hu, Z. Tong, G. Chen, X. Zhan and Z. Hu, Int. J. Environ. Sci. Technol., 2013, 11, 785–794.
10. B. Li, X. Zhu, K. Hu, Y. Li, J. Feng, J. Shi and J. Gu, J. Hazard. Mater., 2016, 302, 57–64.
11. J. W. Jun, M. Tong, B. K. Jung, Z. Hasan, C. Zhong and S. H. Jhung, Chem. – Eur. J., 2015, 21, 347–354.
12. Y. Wang, L. Sun, T. Han, Y. Si and R. Wang, J. Soils Sediments, 2015, 16, 486–496.
13. C. Liu, Y. Chuang, T. Chen, Y. Tian, H. Li, M. Wang and W. Zhang, Environ. Sci. Technol., 2015, 49, 7726–7734.
14. Y. Wang, G. Morin, G. Ona-Nguema and G. E. Brown, Environ. Sci. Technol., 2014, 48, 14282–14290.
15. Y. Wang, G. Morin, G. Ona-Nguema, F. Juillot, F. Guyot, G. Calas and G. E. Brown, Environ. Sci. Technol., 2010, 44, 109–115.
16. Y. Wang, G. Morin, G. Ona-Nguema, F. Juillot, G. Calas and G. E. Brown, Environ. Sci. Technol., 2011, 45, 7258–7266.
17. Y. Zhang, M. Yang, X. Dou, H. He and D. Wang, Environ. Sci. Technol., 2005, 39, 7246–7253.
18. R. Liu, J. Liu, L. Zhang, J. Sun and G. Jiang, J. Mater. Chem. A, 2016, 4, 7606–7614.
19. W. Chen, S. Li, C. Chen and L. Yan, Adv. Mater., 2011, 23, 5679–5683.
20. J. Luo, J. Liu, Z. Zeng, C. F. Ng, L. Ma, H. Zhang, J. Lin, Z. Shen and H. J. Fan, Nano Lett., 2013, 13, 6136–6143.
21. H. Cong, X. Ren, P. Wang and S. Yu, ACS Nano, 2012, 6, 2693–2703.
22. J. H. Kwon, L. D. Wilson and R. Sammynaiken, Materials, 2014, 7, 1880–1898.
23. J. Hu, Z. Tong, G. Chen, X. Zhan and Z. Hu, Int. J. Environ. Sci. Technol., 2014, 11, 785–794.
24. B. M. Voelker and B. Sulzberger, Environ. Sci. Technol., 1996, 30, 1106–1114.
25. T. Wang, L. Zhang, H. Wang, W. Yang, Y. Fu, W. Zhou, W. Yu, K. Xiang, Z. Su, S. Dai and L. Chai, ACS Appl. Mater. Interfaces, 2013, 5, 12449–12459.
26. X. Liu, W. Zhang, Y. Hu and H. Cheng, Microchem. J., 2013, 108, 38–45.
27. Q. Cao, F. Huang, Z. Zhuang and Z. Lin, Nanoscale, 2012, 4, 2423–2430.
28. J. Zhao, Z. Wang, Q. Zhao and B. Xing, Environ. Sci. Technol., 2014, 48, 331–339.
29. J. Saiz, E. Bringas and I. Ortiz, Ind. Eng. Chem. Res., 2014, 53, 18928–18934.
30. N. A. Zubir, C. Yacou, J. Motuzas, X. Zhang and J. C. D. Da Costa, Sci. Rep., 2014, 4, 4594.
31. R. A. D. Y. Peng-Gang, Nanotechnology, 2011, 22, 55705.
32. T. P. Joshi, G. Zhang, W. A. Jefferson, A. V. Perfilev, R. Liu, H. Liu and J. Qu, Chem. Eng. J., 2017, 309, 577–587.
33. K. Liu, S. Zhang, X. Hu, K. Zhang, A. Roy and G. Yu, Environ. Sci. Technol., 2015, 49, 8657–8665.
34. X. Wang, J. Lu and B. Xing, Environ. Sci. Technol., 2008, 42, 3207–3212.
35. X. Xie, Y. Hu and H. Cheng, Water Res., 2016, 89, 59–67.
36. X. Li, J. J. Pignatello, Y. Wang and B. Xing, Environ. Sci. Technol., 2013, 47, 8334–8341.
37. A. Adamescu, I. P. Hamilton and H. A. Al-Abadleh, J. Phys. Chem. A, 2014, 118, 5667–5679.
38. Q. Fang, B. Chen, Y. Lin and Y. Guan, Environ. Sci. Technol., 2014, 48, 279–288.
39. Y. Zhang, M. Yang and X. Huang, Chemosphere, 2003, 51, 945–952.
40. K. Chakarova, N. Drenchev and K. Hadjiivanov, J. Phys. Chem. C, 2012, 116, 17101–17109.
41. X. Zhou, J. Wei, K. Liu, N. Liu and B. Zhou, Langmuir, 2014, 30, 13861–13868.
42. Y. Du, H. Fan, L. Wang, J. Wang, J. Wu and H. Dai, J. Mater. Chem. A, 2013, 1, 7729–7737.
43. Z. Wen, C. Dai, Y. Zhu and Y. Zhang, RSC Adv., 2015, 5, 4058–4068.
44. G. Zhou, D. Wang, L. Yin, N. Li, F. Li and H. Cheng, ACS Nano, 2012, 6, 3214–3223.
45. T. Wang, W. Yang, T. Song, C. Li, L. Zhang, H. Wang and L. Chai, RSC Adv., 2015, 5, 50011–50018.
46. N. J. Singh, S. K. Min, D. Y. Kim and K. S. Kim, J. Chem. Theory Comput., 2009, 5, 515–529.
47. L. Wang, J. Gao, F. Bi, B. Song and C. Liu, J. Phys. Chem. A, 2014, 118, 9140–9147.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7en00758b

---