Mass production of low-boiling point solvent- and water-soluble graphene by simple salt-assisted ball milling

Developing a mass production method for graphene is essential for practical usage of this remarkable material. Direct exfoliation of graphite in a liquid is a promising approach for production of high quality graphene. However, this technique has three huge obstacles to be solved; limitation of solvent, low yield and low quality (i.e., multilayer graphene with a small size). Here, we found that soluble graphite produced by mechanochemical reaction with salts overcomes the above three drawbacks. Soluble graphite was exfoliated into monolayer graphene with more than 10% yield in five minutes of sonication. The modified graphite was easily exfoliated in a low-boiling point solvent such as acetone, alcohol and water without the aid of a surfactant. Molecular simulation revealed that the salt is adsorbed to the active carbon at the graphite edge. In the case of weak acid salts, the original bonding nature between the alkali ion and the base molecule is retained after the reaction. Thus, alkali metals are easily dissociated in a polar solvent, leading to negative charge of graphene, enabling the exfoliation of graphite in low boiling point solvents. The approach proposed here opens up a new door to practical usage of the attractive 2D material.


Ball milling condition
In our experiment, planetary ball milling (P-6, Fritsch) was used to facilitate the mechanical reaction of graphite with salts. Ball milling condition should be carefully selected. We found that impact mode of ball milling is more suited than friction mode for mechanochemical reaction. There are two important factors to fragment graphite: ball size and number of ball. Smaller ball size is preferable, because frequency of impact between balls dramatically increase with decreasing the size of ball. However, the crashing energy per one impact also decrease, and fragmentation of graphite unlikely occurs when the size of ball is smaller than the threshold value. The threshold value we confirmed was in the range of 1 to 5 mm of diameter, depending on the rotation speed. To ensure mechanochemical reactions occur, we choose steel balls with a diameter of 20 mm in most cases. The mode of ball (impact or friction) depends on the fill factor of balls in a container. If five balls are added, balls are rotated inside the container rather than impact. We cannot check the ball movement visually, but estimate the mode by sound during milling. In the case of impact mode, big sound of impact is heard, in contrast small sliding sound is heard for shear sliding. The mode is changed from shear sliding to impact, when we added the balls more than 6. In the case of shear sliding, the graphite might be exfoliated, but solubility of graphite (yield of graphite in LPE process) is worse than the graphite milled at impact mode. Thus, seven balls with 20 mm diameter was usually added into a container to occur impact of balls. These condition (size of ball and filling factor of ball) depends on the machine and volume of container, and should be determined by experimental try and error in the case of planetary ball milling. With respect to the tumbling milling, there is an equation to calculate the critical rotational speed that can avoid the corotation of balls at walls of container: where d m is an inner diameter of container and d B is a diameter of ball.

Characterization
The size of synthesized powder was precisely measured using SEM (JSM6510-LA, JEOL Ltd.).
XRD (X'pert-MPD-OES, PANalytical) was used to measure the crystallite size (L a ), the crystallite thickness (L c ), and the d-space of graphite, using 10 mass% of silicon powder (SRM 640e, NIST) as reference powder for XRD measurement. CuKa radiation was used for measurement. Excitation voltage and electric current were maintained at 45 kV and 40 mA, respectively. Element analysis was performed by EPMA (JXA-8200, JEOL Ltd.). XPS (K-ALPHA, Thermo Scientific) and FT-IR (FTIR-8300, Shimadzu) measurements were also performed to characterize the synthesized powders. For XPS measurement, the surface of the powder was removed by argon etching for 110 s. Quality of synthesized powder and graphene after LPE process was characterized by Raman spectroscopy (NRS-4100, JASCO Co.) using 532 nm laser. The measurement was conducted more than five times and average values of I D /I G were determined. The graphene concentration of the dispersion was measured by optical absorbance at 660 nm using spectrophotometer (ASV11D, Shimadzu). At least three dispersions were checked to obtain average yield of graphene in LPE process. The size of nanosheet was measured using AFM (SPM-9700, Shimadzu) set to dynamic tapping mode. Zeta potential measurements of FLG in water were taken by nano Partica SZ-100 (Horiba, Ltd.). First, 1 ml of centrifuged dispersion was diluted by pH controlled water more than 10 times, then the diluted solution was poured into the cell for zeta potential measurement. The measurement was repeated three times and an average value of zeta potential was determined. Detail method was described below.

XRD analysis
For XRD measurement, we followed to a standard procedure of X-ray diffraction mesuremant on carbon materials 1 . The peak profiles obtained by conventional XRD measurement depend on sample preparation and measurement condition. The key point of the standard process is adding 10 mass% of silicon powder (SRM 640e, NIST). By comparing with both the carbon profile and silicon profile, the d-spacing, L a , and L c values of the graphite sample were determined. CuKa radiation was used for measurement. Excitation voltage and electric current were maintained at 45 kV and 40 mA, respectively.
The d 002 was determined using Bragg's equation below: , (2) . In the case of L c , FWHM B at C(002) and FWHM b at Si(110) were used. L a and L c B b   were calculated by substituting the  value into the equation below: The shape factor, K, in Scherrer's equation is assumed to be 1.

AFM measurement
It is challenging to deposit nanosheets in a dispersion onto substrates without aggregation occurring.
In our process, the dispersions (IPA or IPA/water cosolvent) were diluted with IPA to realize the momentary evaporating of solvent during spray coating. The optical absorbance of the diluent was controlled approximately A = 0.2. Bath sonication was applied for 1 min, then the dispersion was spray-coated onto freshly cleaved mica heated at 180 ˚C by a hot plate. By this method, the graphene without restacking was observed by AFM. It should be noted that we could not deposit graphene on Si and Si/SiO 2 wafers (including 100nm of oxide layer) without agglomeration by above method. This is presumably due to the weak interaction between graphene surface and Si wafers.

Conductivity of solution
In order check dissociation of soluble graphite in water, conductivity of deionized water with soluble graphite was measured. Proper amount of graphite was added into 100 ml of deionized water, then probe-sonication was conducted for 5 min. The electrical resistivity R of the dispersion was measured by bridge circuit method. By this method, high electrical resistivity of deionized water can be precisely measured. The probe with a cell constant C o of 95 m -1 was used for the measurement. The temperature of dispersion was stabilized at 25 ˚C, then electrical resistivity of the dispersion was measured at least twice. The conductivity of dispersion was determined by multiplying cell constant and electrical resistivity.

Molecular simulation
The results of a first-principles molecular simulation of a chemical reaction between a salt molecule (CH 3 COOK and KNO 3 ) and a rectangular-shape graphene fragment with three sides terminated by hydrogen are shown here. The data for Na 2 CO 3 and Na 2 SO 4 are also shown here. The simulation method is explained in the main article. After structural optimization, we found that a -CO 2 , -SO 2 , or -NO 2 base is absorbed on top of one edge carbon atom in "Y"-shape perpendicular to the graphene plane (see Fig. 5a-d for K 2 CO 3 and K 2 SO 4 and Fig. S14a-d for CH 3 COOK and KNO 3 ) with the adsorption energy more than 5 eV; see Table S4. The electrostatic potential felt by each electron is plotted in Fig. 5e,f for K 2 CO 3 and K 2 SO 4 and Fig. S14e,f for CH 3 COOK and KNO 3 together with the value of the Hirshfeld charge. Obviously, the electrostatic potential of weak acid salts (blue region) is much lower than that of strong acid salts (yellow region). As well, the Hirshfeld charge of the Yshape base is negatively much larger for weak acid salts (typically ~ -0.5) than for strong acid salts (typically ~ -0.1; see Table S4. The bond length between the alkali atom ion and the base molecule is shorter in the case of weak acid salts and the alkali atom ion is more strongly bonded to the graphene edge in the case of strong acid salts; see Table S5. Ball milling 500rpm, 30min IPA 100 ml C i =3 g/L Sonication 5 min Centrifugation 1500rpm, 30min          thickness L c and crystallite size L a are determined based a standard procedure of Xray diffraction measurements on carbon materials.   Fig. S15. In the case of weak acid (CO 3 and CH 3 COO), potassium (or sodium) is close to O1, which is the -CO 2 base, indicating that the original bonding nature between the alkali atom ion and the base molecule is kept.
On the other hand, the potassium of strong acid salt (SO 4 and NO 3 ) is close to O2 or C1 atom. It means the alkali atom ion is more strongly bonded to the graphene edge.  Hirshfeld charge of the -NO 2 base -0.124