Adsorptive Removal of Arsenic by Mesoporous Iron Oxide in Aquatic Systems

: In an e ﬀ ort to explore the potential for the removal of arsenic from water, mesoporous iron oxide (MI) was prepared using a sonochemical method. The surface area and average pore size of the MI sample were determined using a Brunauer Emmett Teller (BET) analysis to be 269 m 2 / g and 6.9 nm, respectively. Kinetic experiments revealed that more than 90% of the As(III) and As(V) were adsorbed onto the MI sample within 5 min at 0.1 g / L of MI dosage. The Langmuir equation model suited As(V), whereas As(III) related better to the Freundlich equation model due to di ﬀ erent adsorption mechanisms. The predominant mechanism of As(V) adsorption onto MI is thought to be the electrostatic force between As(V) and MI, whereas As(III) in the solution bound to the adsorbed As(III) on the MI in a way consistent with van der Waals attraction. The removal rate of As(III) and As(V) has the highest adsorption e ﬃ ciency in the pH 5–9 range. The adsorption of As(III) and As(V) was little a ﬀ ected by ionic strength, however the presence of H 4 SiO 40 and PO 43 − signiﬁcantly reduced the arsenic adsorption capacity. Furthermore, the adsorption and regeneration e ﬃ ciency of MI was maintained at around 100% for given adsorption–regeneration cycles.


Introduction
Arsenic (As), a common constituent of the earth's crust, is well-known as a carcinogenic element [1]. Arsenic is introduced into the environment through a combination of natural processes (i.e., weathering reactions, biological activities, and volcanic emissions) and anthropogenic activities [2] and is usually found in the form of inorganic oxyanions. In general, arsenate (As(V)) exists in aerobic conditions, whereas arsenite (As(III)) is dominant in anoxic conditions. As(III) is considerably more mobile and toxic than As(V) [3]. The valence states and chemical forms are strongly dependent on pH, redox potential, and the presence of complexing ions [3,4].
The chronic toxicity of arsenic in drinking water causes various types of cancer and Blackfoot disease [5]. For this reason, the World Health Organization (WHO) and the United States Environment Protection Agency (US EPA) have classified arsenic as a known human carcinogen, and a new maximum contaminant level (MCL) of 10 µg/L for arsenic in drinking water was enforced in 2006 [6,7]. Therefore, the development of an effective arsenic treatment process has been required to meet the tightened MCL. Among the possible treatment processes, adsorption is considered the most cost-effective, the easiest, and the safest process [3,4,8]. These benefits have attracted researchers to use adsorption for arsenic removal from drinking water; to date, various materials such as silica [9],

Materials Characterization
X-ray diffraction (XRD) patterns of MI were obtained from a Rigaku D/MAX Ultima III diffractometer (Japan) using Cu/Kα radiation, and N 2 adsorption/desorption experiments were carried out using a Micromeritics ASAP 2010 volumetric adsorption analyzer (USA). Transmission electron microscopy (TEM) images of the materials were then acquired on a JEOL 2010 electron microscope (Japan) operating at 200 kV. The point of zero charge (PZC) of the MI was determined with an Otsuka Electronics ELS-8000 zeta potential analyzer (Japan).

Batch Experiment
Stock solutions were prepared using sodium arsenate (Na 2 HAsO 4 ·7H 2 O) for As(V) and sodium arsenite (NaAsO 2 ) for As(III) (Sigma Aldrich Chemical Co., St. Louis, MO, USA). Batch adsorptions of As(III) and As(V) were carried out in 50 mL polypropylene copolymer centrifuge tubes containing 3.75 mg of MI in 50 mL of arsenic solution in 0.01 M NaCl at pH 7. The samples were placed on a shaker and mixed continuously for 12 h. The samples in the test tubes were filtered through a 0.45 µm polyvinylidene fluoride (PVDF) filter, and the filtered solution was analyzed for total arsenic using an inductively coupled plasma-optical emission spectrometer (ICP-OES; Perkin Elmer, Waltham, MA, USA). A series of batch isotherm tests were performed for As(III) and As(V) adsorption by varying the initial arsenic concentration (1-9 mg/L) and the temperature (278, 298, 318 K) using 0.075 g/L of MI at pH 7 (total volume: 50 mL). The mixtures were shaken for 12 h at 200 rpm to reach equilibrium. Batch kinetic tests were conducted to determine the rate of arsenic removal by MI. The experiments were initiated by adding 0.05, 0.075, and 0.1 g/L of MI into 50 mL of a solution containing 1 mg/L of As(III) and As(V) at pH 7. The samples were taken out at predetermined times and then analyzed for arsenic after passing through a 0.45 µm PVDF. The rate constants (k d , g/mg min) were calculated using the pseudo-second-order kinetic model, and the intraparticle diffusion model was also used for the analysis of the rate-limiting step of the adsorption. The effect of pH (range 2-12) was studied by adding 0.1 M HCl or 0.1 M NaOH, as needed. The effect of ionic strength on arsenic adsorption was determined using inert electrolyte KCl solutions (0-50 mM). In addition, batch tests were performed to observe the effect of competitive anions using solutions of 1 ppm As(III) and As(V) in 10 mM solutions of NaH 2 PO 4 , Na 2 SiO 3 , Na 2 SO 4 , NaNO 3 , and NaHCO 3 . After a 12-h reaction time, the arsenic concentration was analyzed using ICP-OES. The regeneration test was conducted to estimate the reusability of MI. In these experiments, a NaOH solution (0.1 M) was used as the desorption solution. Figure 1 shows the N 2 adsorption-desorption measurement and pore size distribution of fresh MI. According to the IUPAC classification of adsorption isotherms for gas-solid equilibrium, Figure 1 reveals that MI was classified as Type IV, indicating the pore size of MI was determined to be a mesoporous structure [32]. Moreover, by the modern classification of hysteresis loops, MI has an H-2 loop type [32]. The H-2 loop is broad, with the desorption branch being much steeper than the adsorption branch. It is accepted that this kind of loop occurs when there is a difference in the mechanism between condensation and evaporation and that this occurs in pores with narrow necks and widebodies or when the porous material has an interconnected pore network. BET specific surface area, pore volume, and mean pore size is shown in Figure 1. The pore size distribution confirms the mean pore size and volume of MI is approximately 6.9 nm and 0.44 cm 3 /g, respectively. The BET surface area is estimated to be 269 m 2 /g. In addition, a wide-angle XRD measurement was conducted to confirm the phase structure of MI. As shown in Figure S1, the main diffraction peaks for MI, observed at 2θ, indicate that sonochemically synthesized MI is hematite (α-Fe 2 O 3 ). In the TEM image ( Figure S2), the bright area indicates the pores, and the dark region is the solid phase, and the MI has an irregular shape with diameters in the 0.5-1 µm range. Uniformity in the pore structure is hardly visible for mesostructures, and the structure is more likely of a wormhole-like morphology. The point of zero charge (PZC) of MI was determined to be around 8.4, which is similar to hematite [33,34], and the zeta potential of MI at pH 7 was measured to be 9.43 mV.

Characterization of MI
Water 2020, 12, x FOR PEER REVIEW 4 of 12 hardly visible for mesostructures, and the structure is more likely of a wormhole-like morphology. The point of zero charge (PZC) of MI was determined to be around 8.4, which is similar to hematite [33,34], and the zeta potential of MI at pH 7 was measured to be 9.43 mV.

Adsorption Isotherms
Langmuir and Freundlich isotherms were examined to analyze the adsorption characteristics of the adsorbent. The linear forms of these equations are displayed as Equation (1) (Langmuir model) and Equation (2) (Freundlich model): where qm (mg/g) is the maximum adsorption capacity, qe (mg/g) is the amount of adsorbed arsenic, Ce (mg/L) is the equilibrium arsenic concentration, KF and n are Freundlich constants, and KL (L/mg) is the Langmuir constant. The linear Langmuir and Freundlich isotherms were subsequently fitted to the experimental data ( Figure 2).

Adsorption Isotherms
Langmuir and Freundlich isotherms were examined to analyze the adsorption characteristics of the adsorbent. The linear forms of these equations are displayed as Equation (1) (Langmuir model) and Equation (2) (Freundlich model): where q m (mg/g) is the maximum adsorption capacity, q e (mg/g) is the amount of adsorbed arsenic, C e (mg/L) is the equilibrium arsenic concentration, K F and n are Freundlich constants, and K L (L/mg) is the Langmuir constant. The linear Langmuir and Freundlich isotherms were subsequently fitted to the experimental data ( Figure 2).
Water 2020, 12, x FOR PEER REVIEW 4 of 12 hardly visible for mesostructures, and the structure is more likely of a wormhole-like morphology. The point of zero charge (PZC) of MI was determined to be around 8.4, which is similar to hematite [33,34], and the zeta potential of MI at pH 7 was measured to be 9.43 mV.

Adsorption Isotherms
Langmuir and Freundlich isotherms were examined to analyze the adsorption characteristics of the adsorbent. The linear forms of these equations are displayed as Equation (1) (Langmuir model) and Equation (2) (Freundlich model): where qm (mg/g) is the maximum adsorption capacity, qe (mg/g) is the amount of adsorbed arsenic, Ce (mg/L) is the equilibrium arsenic concentration, KF and n are Freundlich constants, and KL (L/mg) is the Langmuir constant. The linear Langmuir and Freundlich isotherms were subsequently fitted to the experimental data ( Figure 2). The results of fitting the Freundlich and Langmuir equations to isotherm curves are summarized in Table S1. The maximum adsorption capacity of As(III) and As(V) calculated by the Langmuir isotherm was 136.89 mg/g and 31.82 mg/g at 298 K, respectively. Based on the adsorption isotherm analysis, adsorption of As(III) on MI was well fitted to the Freundlich model, whereas the Langmuir model agreed well with the experimental data of As(V) removal. For further comparison of the fitting of experimental data to isotherm models, a non-linear Chi-square analysis (χ 2 ) suggested by Ho.  was also conducted. χ 2 is determined using the following equation (Equation (3)): where q e (mg/g) is the experimental equilibrium capacity, and q e , m (mg/g) is the equilibrium capacity calculated from the models. As Ho et al. suggested, the smaller χ 2 values display a better match to the model [35]. As shown in Table S1, the χ 2 values of the Freundlich isotherm for As(III) adsorption were lower than those of the Langmuir isotherm, indicating that the Freundlich isotherm is the better fitting isotherm for the adsorption of As(III) on MI. The results indicated that As(III) adsorbed onto MI in multilayer and also adsorption increases with increasing the concentration of As(III). In contrast, the χ 2 values of the Langmuir isotherm were lower than the Freundlich isotherm for the As(V) cases; i.e., the Langmuir equation represents a better fit of the experimental data than the Freundlich equation. This indicates that As(V) adsorption could be characterized by a steep initial rise that approaches a plateau attributed to a complete monolayer's formation. When the initial As(III) and As(V) concentration is below 2 mg/L, the adsorption capacities of MI for As(V) are higher than that for As(III), whereas the adsorption capacities of MI for As(III) are higher when the initial arsenic concentration is increased to more than 2 mg/L.
In general, the adsorption capacity of As(V) is higher than As(III) due to electric charge and mobility [36]. In contrast, MI show opposite arsenic adsorption tendency, and a similar trend was reported by Tang et al. when they use ultrafine α-Fe 2 O 3 nanoparticles as an arsenic adsorbent [37]. The different adsorption trends for As(III) and As(V) in this study suggest that different mechanisms outside of the existing adsorption controlled processes-not related to surface charge for arsenic adsorption-might be occurring during As(III) adsorption. In order to identify this process, the surface charge of arsenic adsorbed MI was measured. At pH 7, the surface charge of As(III) and As(V) adsorbed MI were −0.25 mV and −10.95 mV, respectively. In this neutral condition, the MI adsorbed with As(III), and the As(III) has a neutral charge. Therefore, it is posited here that MI adsorbed with As(III) was weakly bound to As(III), which can be explained using the van der Waals attraction model; this, in turn, explains why this data shows that As(III) adsorption follows the Freundlich model. In other words, As(III) was adsorbed to MI via electrostatic force until As(III) occupied the active surface of MI, at which point the As(III) in the solution bound to the adsorbed As(III) on the MI as per van der Waals attraction. Whereas the predominant mechanism of As(V) adsorption onto MI is attributed to the electrostatic force between As(V) and MI, in this case, As(V) was adsorbed onto MI via electrostatic attraction as a monolayer formation, and then electrostatic repulsion occurred between As(V) and As(V) adsorbed MI. This explains why the isotherm experiment results of As(V) adsorption follow the Langmuir model.
Thermodynamic parameters were also calculated based on Equations (4)-(6) and reported in Table S2.
∆G is the Gibbs free energy change, R is the ideal gas constant (4.187 J mol −1 K −1 ), T is the temperature (K), K is the Langmuir or Freundlich isotherm constant, ∆H is the enthalpy change, and ∆S Water 2020, 12, 3147 6 of 12 is the entropy change. As listed in Table S2, the adsorption capacities of As(III) and As(V) are seen to rise with increasing temperatures, possibly due to the activation of more sites on the MI surface at higher temperatures. A positive value of the standard enthalpy change (∆H • ) for As(III) and As(V) adsorption on the MI surface indicates the endothermic nature of the process, a negative value of the free energy indicates the spontaneous nature of the adsorption process.

Kinetics of Arsenic Adsorption
Arsenic adsorption kinetics were examined to investigate the prepared MI's adsorption behavior using different MI dosages with an initial pH of 7. The removal of As(III) and As(V) were then fitted to the pseudo-second-order kinetic model, which has been frequently utilized for the adsorption capacity of mesoporous materials [26,28]. The effect of MI dosages (0.05, 0.075, and 0.1 g/L) on the rate of adsorption using 1 mg/L of As(III) and As(V) are shown in Figures 3a and S3. The adsorption rate was rapid in the first 30 min and lowered considerably as the reaction approached equilibrium. Figure S3 shows that 60% and 70% of As(III) and As(V) were adsorbed within only 5 min, and the performance reached 90% and 100% in 1 h for the 0.1 g/L MI nanoparticle. Moreover, arsenic adsorption rose with an increase in the adsorbent concentration (from 0.05 to 0.1 g/L). This increase in the adsorbent dose corresponds to an increase in the number of active sites and enhancing the arsenic adsorption. Hence, an MI concentration of 0.075 g/L, at which As(III) and As(V) removal was improved with minimum MI concentration, was used in all subsequent experiments unless otherwise specified. The pseudo-second-order kinetic model (Equation (7)) [38] was applied to describe arsenic adsorption to calculate the sorption changes.
where K ad (g/mg min) is the rate constant of the pseudo-second-order kinetic model, q e is the amount of adsorbate sorbed at equilibrium (mg/g), and qt is the amount of adsorbate sorbed at t (min). Note that the pseudo-second-order rate constants (k ad ) and surface area normalized rate constants (k sa ) for arsenic removal by MI were calculated and reported in Table 1. For the initial arsenic concentration of 1 mg/L with 0.05-0.1 g/L of MI doses, the rate constant values were 0.0073-0.1443 g/mg min for As(III) and 0.0054-0.1138 g/mg min for As(V). Comparing the arsenic adsorption capacity with other nanomaterials such as nanoscale zero-valent irons, MI shows similar 1ppm As(V) removal efficiency despite using only a one-fifth of the adsorbent dosage in the literature, which is attributed to the enhanced properties in MI in this study [12,17].  Since the general kinetics analysis could not identify the rate-limiting step of the As(III) and As(V) adsorption process on MI, the intraparticle diffusion model was used to analyze the ratelimiting step of the adsorption. The equation (Equation (8)) given by Weber and Morris can be written as [39]: qt = kdt 1/2 (8)  Since the general kinetics analysis could not identify the rate-limiting step of the As(III) and As(V) adsorption process on MI, the intraparticle diffusion model was used to analyze the rate-limiting step of the adsorption. The equation (Equation (8)) given by Weber and Morris can be written as [39]: qt = k d t 1/2 (8) k d is the intraparticle diffusion rate constant (mg/g min1/2) of the first 30 min adsorption step, estimated from the straight line of q t versus t 1/2 . The well fitted straight lines for the initial 30 min of sorption in Figure 3b indicate that intraparticle diffusion might be the rate-limiting step in the period of the first 30 min of sorption. This initial period of 30 min is the typical time used in the literature for intraparticle diffusion based on the macropore/mesopore diffusion process [40]. The high R 2 values in Table S3 support that the predominant adsorption process occurred in the first 30 min; also, fast arsenic adsorption is attributed to the mesopore structure of MI.

Effect of pH on Arsenic Adsorption
The effect of pH on As(III) and As(V) uptake by MI is presented in Figure 4. The removal rate of As(III) and As(V) was high in the pH range (5-9), and the As(III) removal rate decreased to 56.2% at a pH below 4 or above 10. In the case of As(V), after As(V) adsorption reached equilibrium, the sorption of As(V) sharply decreased at the basic condition. The pH dependency of adsorption can be explained via the ionization of both MI and arsenic. When the pH is above 9.2, H 2 AsO 3− is the predominant As(III) species-while the MI surfaces are also negative, causing electrostatic repulsion. Similarly, this trend of pH dependence was observed during As(V) removal using MI. As(V) has pK1, pK2, and pK3 values of 2.2, 7.08, and 11.5, respectively. In a pH range of 2-7, H 2 AsO 4− and HAsO 4 2− are the predominant species of As(V). Because the predominant As(V) species changes in water, electrostatic attraction and repulsion between MI and As(V) occurred. The decrease in As(V) adsorption is higher than As(III) at a high pH due to the more substantial electrical charge of As(V) compared to As(III). A similar pH dependence trend in the arsenic adsorption on nanosized zero-valent iron (NZVI), hematite, and goethite has also been observed [16,17,39].

Effects of Ionic Strength and Anions on Arsenic Adsorption
To determine the effect of ions on arsenic removal using MI, an investigation of how the ionic strength and competing ions affect the experiment was performed. As shown in Figure 5a, the adsorption of As(III) and As(V) was little affected by different ionic strength (0-50 mM). Figure 5b shows the effect of individual anions (HCO 3− , SO4 2− , NO 3− , H4SiO40, and PO4 3− ) on the adsorption of

Effects of Ionic Strength and Anions on Arsenic Adsorption
To determine the effect of ions on arsenic removal using MI, an investigation of how the ionic strength and competing ions affect the experiment was performed. As shown in Figure 5a, the adsorption of As(III) and As(V) was little affected by different ionic strength (0-50 mM). Figure 5b shows the effect of individual anions (HCO 3− , SO 4 2− , NO 3− , H 4 SiO 40 , and PO 4 3− ) on the adsorption of As(III) and As(V) on MI. The HCO 3− , SO 4 2− , and NO 3− ions had a slight effect on the As(III) uptake in 10 mM concentrations, whereas the same concentration of PO 4 3− and H 4 SiO 40 , reduced the adsorption from 79.6% to 11.9% and 0%, respectively. Similar results were reported by Su and Puls [41]. Adsorption of As(V) was slightly affected by NO 3− and SO 4 2− anions, whereas HCO 3− , PO 4 3− and H 4 SiO 40 decreased the sorption from 91.9% to 46.5%, 7.2%, and 1.3%, respectively. According to the batch experiments results, for As(III), the adsorption maximum and affinity parameters do not statistically differ among the 5mM HCO 3− concentrations, whereas HCO 3− ions exhibit a noticeable influence on the As(V) adsorption onto MI. In the case of anions, anions are well adsorbed on iron oxide adsorbents with positive charge. For anions, they are readily adsorbed onto the iron oxide adsorbents with positive charge via electrostatic attraction. In particular, it was hypothesized that As(V) arsenic is more affected because it has a greater negative charge than As(III). Similary, Suzuki explained that HCO 3− acted as a competitor for adsorption sites for As(V) but not for As(III) [42]. Whereas, in the case of PO 4 3− , not only is the amount of charge affected, but also the structural effect. When the PO 4 3− is present, the phosphate ion has similar physical and structural properties as the arsenic ions because both ions are located in the same group of the periodic table [43]. Moreover, As(V), PO 4 3− and H 4 SiO 40 are all tetrahedral anions; consequently, all anions form inner-sphere complexes with the functional groups at the surfaces of iron oxides [44]. Moreover, in H 4 SiO 40 , a significant portion of silica existed in colloidal form due to low solubility (183 ppm at pH 7, 50 • C) of silica [45]. Consequently, the adsorption capacity of MI might be decreased because of the fouling of the sorbent pores/surface caused by the colloidal silica. Therefore, the phosphate ion and silicate ion act as strong competitors of arsenic adsorption.

Regeneration of Arsenate Adsorbed MI
In Figure 4, the adsorption of As(V) decreased as the pH value increased, which indicates that an adsorbed As(V) ion can be desorbed from MI by alkaline media. In order to prevent the dissolution of mesoporous iron oxide, desorption experiments were conducted using As(V) adsorbed MI with 0.1 M NaOH solution, and the reaction time varied in a range of up to 6 h. As shown in Figure 6, the desorption procedure reached equilibrium within 30 min. More than 60% of the As(V) was desorbed from the MI within 30 min, and no iron ion was detected in the treated solution. In addition, the feasibility of the adsorption-desorption cyclic operation was examined to check out the long-term performance of MI. Figure S4 shows repeated runs of As(V) adsorption in the regenerated MI, with a fresh supply of 3 mg/L As(V) added at the beginning of each cycle. In every cycle, the As(V) was completely adsorbed. In the 0.2 g/L MI dosage, an average of 2.25 ppm As(V) was adsorbed after five

Regeneration of Arsenate Adsorbed MI
In Figure 4, the adsorption of As(V) decreased as the pH value increased, which indicates that an adsorbed As(V) ion can be desorbed from MI by alkaline media. In order to prevent the dissolution of mesoporous iron oxide, desorption experiments were conducted using As(V) adsorbed MI with 0.1 M NaOH solution, and the reaction time varied in a range of up to 6 h. As shown in Figure 6, the desorption procedure reached equilibrium within 30 min. More than 60% of the As(V) was desorbed from the MI within 30 min, and no iron ion was detected in the treated solution. In addition, the feasibility of the adsorption-desorption cyclic operation was examined to check out the long-term performance of MI. Figure S4 shows repeated runs of As(V) adsorption in the regenerated MI, with a fresh supply of 3 mg/L As(V) added at the beginning of each cycle. In every cycle, the As(V) was completely adsorbed. In the 0.2 g/L MI dosage, an average of 2.25 ppm As(V) was adsorbed after five cyclic operations; it was thus noted that MI could undergo repeated adsorption-desorption cycles without a noticeable loss of adsorption activity.
In Figure 4, the adsorption of As(V) decreased as the pH value increased, which indicates that an adsorbed As(V) ion can be desorbed from MI by alkaline media. In order to prevent the dissolution of mesoporous iron oxide, desorption experiments were conducted using As(V) adsorbed MI with 0.1 M NaOH solution, and the reaction time varied in a range of up to 6 h. As shown in Figure 6, the desorption procedure reached equilibrium within 30 min. More than 60% of the As(V) was desorbed from the MI within 30 min, and no iron ion was detected in the treated solution. In addition, the feasibility of the adsorption-desorption cyclic operation was examined to check out the long-term performance of MI. Figure S4 shows repeated runs of As(V) adsorption in the regenerated MI, with a fresh supply of 3 mg/L As(V) added at the beginning of each cycle. In every cycle, the As(V) was completely adsorbed. In the 0.2 g/L MI dosage, an average of 2.25 ppm As(V) was adsorbed after five cyclic operations; it was thus noted that MI could undergo repeated adsorption-desorption cycles without a noticeable loss of adsorption activity.

Conclusions
In this research, sonochemically synthesized mesoporous iron oxide (MI) showed high adsorption capacity for the removal of As(III) and As(V). Adsorption of As(III) and As(V) on MI was well fitted to the pseudo-second-order kinetic model with high rate constants compared to other

Conclusions
In this research, sonochemically synthesized mesoporous iron oxide (MI) showed high adsorption capacity for the removal of As(III) and As(V). Adsorption of As(III) and As(V) on MI was well fitted to the pseudo-second-order kinetic model with high rate constants compared to other types of nanomaterials. Adsorption mechanism of As(III) on MI follows the Freundlich model based on the van der Waals interaction, whereas that of As(V) on MI was well fitted to Langmuir model due to mainly electrostatic interaction between As(V) and MI or As(V) adsorbed MI. Water characteristics such as competing anions and pH affected arsenic adsorption capacity; on the other hand, ionic strength (<50 mg/L KCl) had a slight effect on MI adsorption capacity. Moreover, arsenic adsorbed MI was efficiently regenerated under basic condition, and MI showed good reusability for arsenic removal. This study suggests that MI has a great potential for field application as an arsenic adsorbent in practical and economic feasibility due to the fast adsorption/desorption rate, simple regeneration process, and high adsorbent stability.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/12/11/3147/s1, Table S1: Related parameters for the adsorption of As(III) and As(V) on MI, Table S2: Thermodynamic parameter for adsorption of As(III) and As(V) on MI at different temperatures. (Langmuir and Freundlich isotherm), Table S3: Intraparticle diffusion rate constants for adsorption of As(III) and As(V) on MI at different MI dosage, Figure S1: Wide-angle X-ray diffraction patterns of MI (α-Fe 2 O 3 ). (Reference: JCPDS file, No. 33-0664), Figure S2: Transmission electron microscope images of MI, Figure S3: Adsorption kinetics of As(III) (a) and As(V) (b) based on pseudo second model with various MI concentrations. The experiments were initiated by adding 0.05, 0.075 and 0.1 g/L of MI into solution containing 1 mg/L of As(III) and As(V) at pH 7, Figure S4: Regeneration cycle of As(V) adsorbed MI. The experiments were initiated by adding 0.2 g/L of MI and 3mg/L As(V) was added at the beginning of each cycle. model with various MI concentrations. The experiments were initiated by adding 0.05, 0.075 and 0.1 g/L of MI into solution containing 1 mg/L of As(III) and As(V) at pH 7.