Oxidation of Arsenite by Epoxy Group on Reduced Graphene Oxide/Metal Oxide Composite Materials

Abstract Reduced graphene oxide/metal oxide (rGO/MO) hybrid has been widely used as a catalyst, while dissolved oxygen or radicals are generally recognized as the oxidant. This study finds that the adsorbed arsenite (As(III)) on rGO/MO is oxidized to arsenate (As(V)) in the absence of other oxidants or radicals. The oxidation of As(III) is observed on varying rGO/MOs, including rGO/MOs composited of different types of reduced graphene oxide (rGO) and metal oxide. The epoxy group on rGO acts as the oxidant, evidenced by the significant correlation between the consumption of epoxy group and oxidation of As(III). Meanwhile, metal oxide provides adsorption sites for As(III) during the adsorption–oxidation process. Based on a combination of spectroscopic measurements and computational calculation, a possible pathway for As(III) oxidation by rGO/MO is proposed: the oxygen atom in the epoxy group is bonded to the adsorbed AsIIIO3, which is consequently oxidized to AsVO4. Overall, this study proves the role of rGO/MO as an oxidant, which opens a new perspective on future studies using rGO/MO as a catalyst for the oxidation reaction.


Contents:
Figure S1 on Page S2: Synthesis of rGO/MO; Text S1 on Page S4: Results of characterizations for rGO/LOa; Figure S2 on Page S5: Electron microscope images; Text S2, Figure S3, and Table S1 on Page S6: Results of Raman analysis; Figure S4 on Page S7: Results of XRD; Figure S5 on Page S8: Correlation analysis of μ-XRF results; Text S3 and Figure S6 on Page S9: Results of EPR for radical tests; Figure S7 on Page S10: First derivative of XANES analysis; Text S4 and Figure S8 on Page S11-S13: Reaction kinetics analysis; Figure S9 on Page S14: 2D-COS analysis of FTIR spectra; Figure S10 on Page S15 and Table S2 on Page S16: C1s XPS analysis; Figure S11 on Page S17: XANES LCF fitting; Table S3 on Page S18: A summary of XANES LCF fitting and XPS fitting; Text S5 on Page S19-20 and Table S4 on Page S21: EXAFS fitting.

Synthesis of graphene oxides
Regarding the Hummers' method, 23 mL H2SO4 was added into 1 g graphite flake and 0.5 g NaNO3 in a 250 mL three-neck flask with a stir bar. After the mixture was cooled in an ice bath for 20 min, 3 g KMnO4 was slowly added and the mixture was maintained in the ice bath for another 2 h. The mixture was heated to 35 o C in the water bath and kept for 1 h. Then, the synthesized viscous dispersion was slowly poured into 250 mL deionized (DI) water (resistivity > 18.2 MΩ·cm). A 30% H2O2 solution (10 mL) was slowly added into the mixture. The obtained mixture was then filtered using the PTFE membrane with a 0.22 μm pore size and washed in the sequence by 1 L DI water, 1 L 10% HCl, and 2 L DI water. The resulting filtered cake was freeze-dried under vacuum condition.
Regarding the Daniela method, a mixed solution of H2SO4 (180 mL) and H3PO4 (20 mL) was added slowly into 1.5 g graphite flake and 9 g KMnO4 in a three-neck flask.
This mixed solution was then heated to 50 o C and kept mixing for 12 h. The flask was then cooled down to room temperature and the resulting mixture was poured onto 200 g ice. The synthesized mixture was then filtered, washed, and dried using the same procedures as those in the Hummers' method.

Synthesis of different types of rGO/MO
To synthesize rGO/LOa, 0.04 g GO synthesized using the Daniela method was added into 150 mL DI water in a three-neck flask. A solution of 1.75 g La(NO3)3•6H2O in 25 mL DI water (La-solution) was added into the GO solution. The pH was then adjusted to 10 using the ammonia solution (NH3•H2O), and 100 µL N2H4 was added into the solution. The solution was then heated to 90 o C using a water bath. After 4 h mixing, the mixture was filtered and washed using DI water until the conductivity below 100 µS cm -1 . The separated solids were freeze-dried under vacuum. The same procedure was also used to synthesize other rGO/MOs, in which 0.97 g Ti(SO4)2 and 2.70 g Al2(SO4)3 were used to synthesize the rGO/TO and rGO/AO, respectively.
For rGO/LOb and rGO/LOc, the GO fabricated by the Hummers method was used.
All reagents were obtained from Sinopharm, China.
The electron microscope image ( Figure S1-a) shows that the rGO films are agglomerated, agreeing well with previous studies. [1] The pristine La(OH)3 is in the shape of nanowire ( Figure S1-b), which changed to nanorod after being deposited on rGO ( Figure S1-c), suggesting that rGO inhibited the growth of La(OH)3. The Raman analysis also confirmed the crystalline change of La(OH)3 after being composited with rGO ( Figure S3, Table S1, and Text S2). The major crystal of rGO/LOa is La(OH)3, as evidenced by the X-ray diffraction pattern shown in Figure S4. In addition, the co-  The Raman analysis was employed to further characterize the properties of these three materials ( Figure S3). The peaks observed on Raman spectra were assigned and summarized in Table S1. The peaks of D-band, G-band, 2D-band, and S3 agreed well with those of the rGO reported in the previous studies. [2] The ratio of peak intensities of D-and G-band (ID/IG) is 1.73-1.76, indicating that GO was reduced to rGO. [2b] The peaks on Raman spectra of La(OH)3 are observed at 281, 339, 452, 610, and 1071 cm -1 , which also agrees well with the previous reports. [3] However, it is interesting that the La(OH)3 deposited on rGO shows new peaks at 396 and 855 cm -1 . This difference is mainly due to the change in the ratio of the crystalline faces, which suggest the chemical composite of rGO and La(OH)3, as is observed from the FESEM images.   Figure S4. The chemical 5, 5-dimethyl-1-pyrroline-N-oxide (DMPO) was used to capture the radicals with 5 g/L rGO/LOa in water under dark condition, which is the same condition used for adsorption and IR experiments. The solution containing 1 mM Co(II) solution, 100 mM H2O2, and 2% dimethyl sulfoxide (DMSO) was used to produce O2·radical, while the solution containing 1 mM Fe(II) and 100 mM H2O2 was used to produce ·OH.
The results showed that no radical was detected in solutions containing rGO/LOa, excluding the possibility that radical would oxidize As(III) to As(V) during the adsorption process.  Figure SX). Zero order (Eq 1), pseudo first order (Eq. 2), and pseudo second order (Eq 3) models were used to describe the kinetics for the reactions: where qe and qt are the IR absorbance of As-O band (highest peak at 680-900 cm -1 ) at equilibrium and at any time t (min), respectively. The k0 (abs·min -1 ), k1 (min -1 ) and k2 (abs -1 min -1 ) are the rate constants for the zero order, pseudo first order, and pseudo second order models, respectively.
To fit the zero order model, the data were analyzed by linear regression using Eq 1; whereas the data were analyzed by linear regression using Eqs 4 and 5 to fit the pseudo first order and pseudo second order models, respectively. The fitting results are shown in Figure S8, as well as the fitting coefficients and rate constants. Obviously, the increasing of As-O band for As(III) on rGO/MOs follows S12 zero order model (R 2 = 0.961-0.996) better than pseudo first order (R 2 = 0.839~0.961) or pseudo second order (R 2 = 0.186~0.957) models. On the other hand, the increasing of As-O band of As(V) adsorption on rGO/LOb follows pseudo second order (R 2 = 0.956) and pseudo second order (R 2 = 0.929) models better than zero order (R 2 = 0.897) model. This difference further confirmed the different mechanisms of the appearance of the As-O band between As(III) and As(V) onto rGO/MOs: the As-O bands of As(III) onto rGO/MOs resulted from the adsorption-oxidation whereas the As-O bands of As(V) onto rGO/MOs resulted from the adsorption only. S13 Figure S8. Reaction kinetics of As(III) adsorption-oxidation on rGO/TO (a), rGO/AO (b), rGO/LOa (c), rGO/LOb (d), rGO/LOc (e), and rGO/LOd (f), as well as As(V) adsorption on rGO/LOb (g); the absorbance data prior to equilibrium were derived from the in-situ FTIR spectra (Figure 2). Inset tables show the fitting coefficient (R 2 ) and the fitted rate constants. S14 Figure S9. Synchronous 2D-COS maps for As(III) adsorption on rGO/LOb. The original FTIR spectra are shown in Figure 3-d.   Figure S11. XANES spectra of As(III) adsorption on rGO/LOa, rGO/LOb, rGO/LOc, and rGO/LOd, as well as rGO/LOb-2, rGO/LOb-3, rGO/LOb-4, and rGO/LOb-5.
Duplicate samples were analyzed for each material. S18

Text S5. EXAFS data analysis
The extended X-ray absorption fine structure (EXAFS) analysis was performed using the ATHENA and ARTEMIS programs in the Demeter computer package. [5] The analytical procedure was similar to our previous studies. [6] The raw data measured in intensities were converted to μ(E), and averaged spectra were used in the analysis. The EXAFS signal χ(k) was extracted from the measured data using the AUTOBK algorithm [7] where k is the photoelectron wave number. The primary quantity for EXAFS is χ(k), which is the oscillations as a function of photoelectron wave number.
χ(k) is weighted by k 3 to account for the dampening of oscillations with increasing k.
The different frequencies in the oscillations in χ(k) correspond to different near neighbor coordination shells which can be described and modeled according to the where f(k) and δ(k) represent the photoelectron backscattering amplitude and phase shift, respectively, N is the number of neighboring atoms, R is the distance to the neighboring atom, and σ 2 is the Debye-Waller factor representing the disorder in the neighbor distance. The k 3 weighted EXAFS spectrum in k-space (Å -1 ) is Fourier transformed (FT) in R-space (Å). The experimental spectra were fitted with single-scattering theoretical phase-shift and amplitude functions calculated with the ab initio computer code FEFF6 [8] using atomic clusters generated from the crystal structures of LaAsO4 (ICSD #415338) and AlAsO4 (ICSD #201774) for different samples. The many-body amplitude reduction factor (S0 2 ) was established as 0.95~1.05 by isolating and fitting the first-shell of As-O. The parameters such as interatomic distance (R), coordination number (CN), the difference in threshold energy (ΔE0) and the Debye-Waller factor (σ 2 ) were first established with reasonable guesses and then fitted in R-space. The error in the overall fits was determined by the R-factor, called the goodness-of-fit parameter. It is defined as R-factor = Σ(χdata-χfit) 2 /Σ(χdata) 2 . Good fits occur for R-factor < 0.05.