Complex Roles of Solution Chemistry on Graphene Oxide Coagulation onto Titanium Dioxide: Batch Experiments, Spectroscopy Analysis and Theoretical Calculation

Although graphene oxide (GO) has been used in multidisciplinary areas due to its excellent physicochemical properties, its environmental behavior and fate are still largely unclear. In this study, batch experiments, spectroscopy analysis and theoretical calculations were addressed to promote a more comprehensive understanding toward the coagulation behavior of GO onto TiO2 under various environmental conditions (pH, co-existing ions, temperature, etc.). The results indicated that neutral pH was beneficial to the removal of GO due to the electrostatic interaction. The presence of cations accelerated GO coagulation significantly owing to the influence of electrical double layer compression. On the contrary, the presence of anions improved the stability of GO primarily because of electrostatic repulsion and steric hindrance. Results of XRD, FTIR and XPS analysis indicated that the coagulation of GO on TiO2 was mainly dominated by electrostatic interactions and hydrogen bonds, which were further evidenced by DFT calculations. The high binding energy further indicated the stability of GO + TiO2 system, suggesting that TiO2 can be used as an effective coagulant for the efficient elimination and coagulation of GO from aqueous solutions. These findings might likely lead to a better understanding of the migration and transformation of carbon nanomaterials in the natural environment.

disturbance in zebrafish larvae. Tu et al. 14 demonstrated that GO could induce cell damage in E. coli by extracting phospholipid molecules from the outer cell membrane. Liao et al. 15 reported that GO showed the greater hemolytic activity of hemoglobin from suspended red blood cells as compared to the aggregated GO. Hence, the understanding of the physical and chemical behavior of GO in the natural environment is significantly to evaluate its toxicity to human beings. From the literature survey, the efficient way to lower the GO concentration in water is the coagulation of GO, which forms large agglomerates that can be easily separated by aggregation, filtration or centrifugation 12,15,16 . Thereby, the coagulation kinetics and stability of GO have been extensively investigated in the literatures 4,[17][18][19] , and the results showed that the stability of GO in aqueous solution was strongly influenced by ionic strength and pH values. In our group, the aggregation behaviors of GO on Al 2 O 3 and layered double hydroxides (LDHs) indicated that the stability of GO was strongly dependent on solution conditions and the coagulation was mainly dominated by the electrostatic interaction and hydrogen bonds 12,16,20 . Similarly, some works also reported that the solution conditions affected the fate and transport of GO in porous media 4,[21][22][23] . However, to the best of our knowledge, the coagulation behavior of GO on titanium dioxide (TiO 2 ) under complicated aqueous environmental conditions have not been reported yet, especially the study from theoretical calculations.
As a multifunctional material, TiO 2 has already been proved to have outstanding performances in photocatalytic [24][25][26] and energy areas [27][28][29] . It has also been widely served as an environmental friendly model mineral for the removal of environmental contaminants due to its well-known properties such as high chemical and structural stability, abundant sources, low solubility at the whole pH range and close to neutral pH for point of zero charge 30 . Particularly, the composites of TiO 2 and carbon-based nanomaterials are currently recognized as highly effective adsorbent and catalyst in the purification of wastewater 24,[31][32][33] . Moon et al. 34 prepared the rGO-TiO 2 composites based on pH-induced aggregation for the efficient photocatalytic oxidation of arsenite. Wang et al. 35 observed the accelerated sedimentation of TiO 2 -GO and attributed it to the aggregation of TiO 2 -GO resulted from the electrostatic attraction between TiO 2 and GO. These studies indicated that TiO 2 may be suitable materials for GO coagulation from solution to solid particles. However, systematic studies on the interaction mechanism between TiO 2 and GO over a broad range of conditions have not been reported heretofore. In addition, spectroscopic evidence and theoretical calculations are crucial to understand the interaction mechanism at molecular level, which is significant to evaluate the migration behavior of GO in natural environment. Such knowledge is significant to better assessment the environmental behavior of toxic GO nanomaterials and to provide basis for further investigation of carbon-based nanomaterials.
The objectives of the current study are: (1) to study the influence of pH, co-existing ions, contact time, temperature and TiO 2 contents on GO coagulation onto TiO 2 ; (2) to characterize the microscopic surface changes before and after GO deposition on TiO 2 using scanning electron microscopy (SEM), X-ray diffraction (XRD), Fourier transformed infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS); (3) to further confirm the interaction mechanisms of GO with TiO 2 by density functional theory (DFT) calculations. It is a highlight to understand the coagulation behavior of GO onto TiO 2 through the investigation of macroscopic experiments, spectroscopy analysis and theoretical calculations. Furthermore, the coagulation behavior offers new insight into the interaction of GO with TiO 2 , which can further understand the behavior of carbon-based materials at solid/ water interfaces in natural environment.

Results and Discussion
Influence of solution pH. The coagulation behaviors of GO in the presence and absence of TiO 2 at different pH values were shown in Fig. 1. The concentration of Ti ions in the solution was analyzed using an inductively coupled plasma mission spectroscopy (ICPE-9000, Shimadzu). It was found that Ti was not detected at 3.0 < pH < 11.0, indicating that TiO 2 owned a good chemical and structural stability. In the absence of TiO 2 , it was clear that the deposition of GO was negligible (Fig. 1A), suggesting that GO was very stable in aquatic solution, which was also evidenced from the visually express in Fig. 1B. The results demonstrated that the GO suspension was very stable in a wide pH range, which was consistent with the literature 17 . However, after adding TiO 2 to GO suspension, one can see that the deposition of GO increased sharply from 5% to 80% as the pH increased from 3.0 to 7.0, and then decreased greatly from 80% to 6% in the pH range of 7.0-11.0. To provide more intuitive evidence about the coagulation process of GO, the reaction photographs of GO with TiO 2 as a function of pH were also presented in Fig. 1B. At low and high pH values, the suspension was homogeneous with a light claybank color, suggesting the good stability of GO suspension. Apparent precipitation was formed at neutral pH, suggesting the strongly deposition of GO onto TiO 2 at neutral pH values. Similarly, the coagulation of GO onto Al 2 O 3 and LDHs was also dependent on pH due to the electrostatic interaction, hydrogen bond and Lewis acid-base interaction 12,16,20 . According to the zeta potential analysis ( Figure S2), the point of zero charge (pH PZC = 6.0) of TiO 2 indicated that the surface of TiO 2 was mainly positively charged at pH < 6.0, whereas the GO was negatively charged at the whole pH. Therefore, the enhanced coagulation of GO onto TiO 2 at pH 2.0-6.0 can be attributed to electrostatic attraction between negatively charged GO and positively charged TiO 2 . The high level deposition of GO at pH 6.0-8.0 was ascribed to the strong chemical interaction and electrostatic interaction 16,20 . At pH > 8.0, deprotonation of carboxylic groups was found to play a vital effect on the stability of GO 16,18,36 . Furthermore, the concentration of OH − in aqueous solution increased with pH increasing, which can compete with GO for interaction with the functional groups on TiO 2 . In consequence, the negatively charged GO was difficult to be attached to the negatively charged TiO 2 at high pH values due to the strong electrostatic repulsion, which resulted in the stable dispersion of GO in aqueous solutions.
Influence of cations. The common environmental cations (Na + , K + , Ca 2+ and Mg 2+ ) were introduced to investigate the coagulation behavior of GO onto TiO 2 . Figure 2A visually expressed the influence of cation types and concentrations on the coagulation of GO. From Fig. 2A, one can see that GO was homogeneously dispersed in low cation concentration and visible precipitation was formed as the cation concentration increased to a critical value (e.g. 20 mM Na + , 10 mM K + , 0.2 mM Ca 2+ and 0.6 mM Mg 2+ ). In order to supply more quantitative knowledge about the deposition behaviors of GO, the equilibrium concentration (C e ) of GO in supernatant at different electrolyte concentrations (NaCl, KCl, CaCl 2 , and MgCl 2 ) were shown in Fig. 2B. GO was stable at relatively low concentrations (e.g. 0.01 to 1.0 mM) for monovalent cations and the difference between Na + and K + was negligible. Above 1.0 mM, the obvious differences among Na + and K + were observed, and the removal percentage of GO increased notably with increasing Na + or K + concentration. The maximum removal percentage of 51% in the presence of Na + and that of 90% in the presence of K + were found. Compared to monovalent cations, the significant deposition of GO was observed in the presence of Ca 2+ and Mg 2+ ions, i.e., as low as 0.2 mM for Ca 2+ and 0.4 mM for Mg 2+ . The destabilization of GO in the presence of different kinds of cations were in the sequence of Ca 2+ > Mg 2+ > > K + > Na + . However, Schulze-Hardy rule showed the similar deposition effect of the same valent cations 18 , which was a little different to the results of this work. This may be attributed to the different interaction properties of cations with GO, and thereby resulted in the more complicated coagulation properties of GO on TiO 2 particles. Wu et al. 18 also showed that the aggregating/destabilizing capacity of divalent cations were higher than monovalent cations and the reason was that divalent cations can interact with the functional groups on GO surfaces, particularly at the edges. The concentrations of Na + and K + in most natural water environment are less than 10 mM, therefore GO will be very stable and have strong migration potential in aqueous solution if no other cations are present in the natural water 37 . While, Ca 2+ and Mg 2+ concentrations in natural aquatic environments often over 0.1 mM, the common environmental divalent cations are more aggressive in GO destabilization, and then results in the deposition of GO in natural water conditions 18,22 . This behavior may be due to the strong binding capacity of divalent cations to functional groups of GO 38 .

Influence of anions.
To further understand the interaction mechanisms of GO with TiO 2 under different solution chemistry conditions, the coagulation process under different concentrations of co-existing anions were shown in Fig. 3. Figure 3A revealed a direct residual GO concentrations in the supernatant after reaction with  4 2− and SiO 3 2− ) had little influence on GO deposition, the minor difference was attributed to their negligible difference on Na loadings. The anion-π interactions between GO and anions occurred as GO can act as an electron acceptor for anion sorption 39 . Shi et al. 40 demonstrated that the anion-π bond was unpredictable strong using a computational method on the basis of density functional theory. Furthermore, Hao et al. 41 reported that 7,7,8,8-tetracyanoquinodimethane had strong π -π stacking interactions with graphene which can effectively prevent the inter-or intra-π -π stacking of graphene, leading to the dispersion of functionalized graphene. Therefore, the negligible effect of anions on GO deposition can be attributed to the strong anion-π interactions which prevented the agglomeration of GO and leading to the excellent water dispersibility. Besides, the anions may be adsorbed on the positively charged surface of TiO 2 through electrostatic attraction, and the surface-adsorbed anions blocked parts of active sites on TiO 2 surface and thus prevented the coagulation of GO due to the steric effects and enhanced electrostatic repulsion.
In the presence of anions, the coagulation of GO can be attributed to a balance between the anions (NO 3 − , Cl − , SO 4 2− and SiO 3 2− ) and Na + . The presence of anions resulted in the more negative surface charge of GO and stabilized GO in solution, while the increase of Na + concentration leaded to the deposition of GO due to electrostatic attraction 12 Figure 4B showed the removal of GO from aqueous solutions by TiO 2 at pH 5.0 ± 0.1 as a function of contact time. The deposition of GO increased quickly with the increase of contact time and reached saturation after 10 hours of contact time. At initial contact time, large amount of functional groups on the surfaces of TiO 2 were available for the binding of GO, which resulted in the quick uptake of GO to TiO 2 surfaces. More aggregation sites and functional groups on TiO 2 were available for the coagulation of GO at the initial contact time, and the deposition of GO on TiO 2 was easily and quickly. With increasing contact time, the available sites were occupied by GO and thereby the uptake of GO became slow. The fast aggregation velocity suggested that strong chemisorption devoted to the deposition of GO onto TiO 2 , which was significant for the application of TiO 2 to remove GO from aqueous solutions in natural environment.
To further gain insight into the coagulation behavior of GO in aqueous solution, the removal percentage of GO at different temperatures were shown in Fig. 4C. It was clear that the temperature played an important role on the deposition of GO and distinct precipitate was formed at T > 50 °C ( Figure S3). According to Fig. 4C, the removal percentage of GO on TiO 2 increased from 25% to 99% with the temperature increased from 20 to 60 °C, indicating that higher temperature was benefical for the aggregation of GO.
The removal percentage of GO on different coagulants were shown in Fig. 4D. One can see that TiO 2 had similar deposition capacity with ZnO and MgO, however it was clearly to see that there were some suspended solids on the bottom of MgO and on the top of ZnO ( Figure S4). The results indicated that the deposition of GO to TiO 2 was more stable than to ZnO and MgO. In addition, it was interesting to notice that TiO 2 has higher deposition capacity as compared with natural clay materials (e.g. bentonite and diatomite). On the basis of aforementioned analysis, it is clearly that TiO 2 can be potentially used as a cost-effective coagulant for the efficient elimination of GO from aqueous solutions, which could efficiently decrease the potential toxicity of GO in the natural environment.  24 . The diffraction peaks of TiO 2 shifted slightly after GO coagulation, indicating that small crystal sized TiO 2 was formed, similar results were also obtained in hybrid TiO 2 @rGO 28 . However, the representative (002) reflection at 2θ = 11.4° of GO was imperceptible, implying a decreased layer-stacking regularity and a highly disordered overlay of individual GO nanosheets were formed in the TiO 2 architecture after GO aggregation 31,44 . Furthermore, from the SEM images ( Figure S5), it was obvious that the surface of TiO 2 was composed of stacked GO nanosheets, revealing that GO had been deposited on the surface of TiO 2 .
As can be seen from the FTIR spectra in Fig. 5B, the GO showed various adsorption bands for water -OH stretching (3400 cm −1 ), carboxylates or ketones C = O stretching (1730 cm −1 ), water -OH bending and C = C stretching (1627 cm −1 ), alcoholic C-OH bending (1400 cm −1 ), epoxide C-O-C (1230 cm −1 ) and C-O stretching (1057 cm −1 ) 26,45 . The fundamental vibrations of TiO 2 appeared at 400-900 cm −1 which were ascribed to the stretching vibrations of Ti-O and Ti-O-Ti bonds, the peak at 3420 cm −1 was due to the stretching of the hydroxyl group 46,47 . The sharp peak at 1432 cm −1 was a characteristic band of hydroxyl group deformation vibration 45   behavior had also been reported for the anchoring of TiO 2 nanoparticles on graphene nanosheets because of the strong chemically bond between GO and TiO 2 49 . The FTIR analysis provided indirect evidence that GO was deposited on the surface of TiO 2 through chemisorption and hydrogen bond.
The interaction mechanisms of GO with TiO 2 were further investigated by XPS. The chemical properties of different elements in TiO 2 , GO and TiO 2 + GO were shown in Fig. 5C-F. The survey spectra (Fig. 5C) showed clearly the existence of C, O in GO, Ti, O in TiO 2 and C, Ti, O in TiO 2 + GO. In addition, the binding energy values of Ti2p and O1s of TiO 2 + GO remarkably shifted to higher wavelength as compared with TiO 2 , which was presumably due to strong chemical interaction between TiO 2 and GO 16 . Similarly, Razzaq et al. 50 observed that the Ti2p and O1s peaks of rGO + TiO 2 shifted to higher binding energies as compared to pure TiO 2 , and they proposed that the drainage of electrons from Ti to rGO and the formation of bonds between TiO 2 with rGO. The high resolution of O 1s spectra were shown in Fig. 5D-F. The O1s spectrum of GO (Fig. 5D) can be divided into three DFT calculation. The interaction mechanism of GO onto TiO 2 was further evidenced by the DFT calculations. The Vienna ab initio simulation package (VASP) (version 5.3.5) was implemented to perform the geometric optimization and static total energy calculations for the coagulation models [52][53][54] . Computational details were summarized in SI. The optimized structure of GO deposition on TiO 2 was shown in Fig. 6. Owing to the oxygen-containing functional groups of GO plane, the minimum Ti(TiO 2 )-O(GO) distance was 2.75 Å. The binding energy (E b ) (Table S1) was calculated as the following: , where E GO , E TiO 2 and + E GO TiO 2 represent energies of the coagulation GO, TiO 2 and the hybrid TiO 2 + GO system, respectively. A high positive E b between GO and TiO 2 (5.79 eV) suggested that the TiO 2 + GO system was stable and the chemisorption was probably the main coagulation mechanism of GO to TiO 2 , complementing to the physisorption behavior 16 , which indicated that strong interaction existed between GO and TiO 2 , and TiO 2 was an effective coagulant for the elimination of GO from natural environment. The calculated charge density distribution and the projected density of states (PDOS) of the hybrid TiO 2 + GO system were shown in Fig. 7. As shown in Fig. 7A, the three dimensional charge density difference plot with an isosurface value of 10 −5 e Å 3 was obtained by subtracting the calculated electronic charges of the individual GO and TiO 2 (101) from that of TiO 2 + GO. The purple and blue bubbles represented positive and negative charges, respectively. It can be clearly seen that the electrostatic interaction was formed between GO and TiO 2 . With further Bader charge analysis 55 , we discovered that there was an average charge transfer of around 0.06 e from GO to TiO 2 (101) facets. Other significant evidence of the notable charge transference between GO and the TiO 2 (101) facets was the comparison between the PDOS plots for GO, TiO 2 and TiO 2 + GO. As shown in Fig. 7B(c), from − 7 to − 2 eV range, there was an obvious hybridization between C, O and Ti orbitals. Because of the interaction and redistribution, the peaks of O1 and C atoms in TiO 2 + GO system almost disappeared when compared with GO. Meanwhile, the peaks of O2 and Ti atoms in TiO 2 + GO system shifted to lower energy levels compared to pure TiO 2 . The theoretical calculation suggested the presence of strong interactions between TiO 2 and GO. Comprehensive consideration of the results of XRD, FTIR, XPS and theoretical calculations, it proved that chemical bonds, hydrogen bonds and electrostatic interactions dominated the coagulation of GO on the surface of TiO 2 from aqueous solutions.
In conclusion, this paper is the first study to investigate the coagulation behavior of GO onto TiO 2 under different environmental solution conditions. Electrostatic interaction is found to play a key role in GO removal by TiO 2 when pH changes. Divalent cations are more effective than monovalent cations in aggregating/destabilizing GO suspensions through the interaction with oxygen-containing functional groups on GO surfaces. GO remains highly stable under different anions due to electrostatic repulsion and steric hindrance. The spectroscopy analysis and DFT calculations further evidence the electrostatic interactions and hydrogen bonds between GO and TiO 2 . The abovementioned insights into the coagulation of GO onto TiO 2 under different solution chemistry conditions are crucial toward understanding the long-term transport and fate of GO.

Methods and Materials
Materials and characterization. GO was synthesized from flake graphite (48 μ m, 99.95% purity) by using the modified Hummers' method 56 . More detailed procedures about the preparation of GO were supplied in Supporting Information (SI). Milli-Q water was used in all experiments. The TiO 2 before and after GO coagulation were characterized by the scanning electron microscopic (SEM), X-ray diffraction (XRD), Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS). More detailed characterization processes were described in SI.

GO Coagulation Experiments.
The coagulation experiments were accomplished in a series of 20 mL vials equipped with Teflon-lined screw caps at 25 ± 1 °C by using batch technique. A certain amount of TiO 2 (0.1 g/L), GO stock suspension (25 mg/L) and the background electrolytes (NaCl, KCl, MgCl 2 , CaCl 2 , NaNO 3 , Na 2 SiO 3 or Na 2 SO 4 ) (0-20 mM) were added to the vials to obtain the required concentrations of different components, and the samples were left undisturbed on a flat surface for 24 h to allow for the complete settlement of TiO 2 and the large GO aggregates. The desired pH was adjusted by adding negligible volumes of 0.01 or 0.1 M HCl and NaOH. The detail experimental process was described in SI. The concentration of GO was analyzed by ultraviolet-visible spectrophotometer (UV-2550, PerkinElmer) at wavelength of 230 nm ( Figure S1). All experimental data were obtained by the average values of triple parallel samples and the error bars were within ± 5%.