Ultrasonic degradation has the advantages of low cost and simple operation (Li et al., 2017), but long-term usage of ultrasonic graphene causes defects, and layer number renders the lamella radial size difficult to control (Khan et al., 2010) and enlarges the scale and the ultrasonic generator (Tian et al., 2017). The traditional ultrasonic degradation process is no longer adopted in present experimental studies and industrial production, and many studies aimed for improvement and innovation (Moorthy et al., 2015; Abo-Zahhad et al., 2016; Li et al., 2017; Narayan et al., 2017; Cai X. Z. et al., 2018; Gai et al., 2018). Gao et al. (2017) combined an ultrasonic generator with a supercritical reactor (Figure 2) by using ultrasonic degradation and supercritical CO₂ (scCO₂) excellent penetration of air to natural graphite in the scCO₂, H₂O, and ethanol system. High-quality graphene (>50%) was obtained in the mixed system. The method for industrial production provides a green, fast, and scalable way of graphene dispersion. Phiri et al. (2017) improved the traditional process of graphene preparation by ultrasonically exfoliating graphite in water medium. In the process of ultrasonic exfoliation, graphene was exfoliated and dispersed by means of a high shear colloidal mixer for 2 h to obtain a graphene dispersion with a concentration of up to 1.1 mg/ml. The graphene fractions prepared by ultrasonic peeling graphite for 100 h were separated by traditional process. Graphene dispersions of the same quality can be prepared in 0.5 h with the improved process. The graphene prepared by the improved process is analyzed by AFM and Raman. The dispersions contain plenty of defect-free and unmodified graphene layers. Hadi et al. (2018) mixed magnetic nanoparticles (Fe₃O₄) and graphite evenly during the preparation of graphene dispersion through the liquid-phase peeling of graphite. During ultrasonic exfoliation, the intense shear force between graphite nanosheets promotes the exfoliation of graphene, greatly shortens the ultrasonic time, and avoids long-term ultrasound bringing defects to graphene surface.

Because of the different parameters of the ultrasonic exfoliation process (e.g., time, frequency, power, and shape of the inner cavity), the experimental results are greatly affected (Río et al., 2017). At present, in the mechanical peeling process (including shearing and ball milling) (Deng et al., 2016; Liu Y. J. et al., 2018), the combination of shear and ball milling exfoliation process has been identified as the trend of replacing the ultrasonic peeling process in the future (Tour, 2014; Bonaccorso et al., 2016). Novoselov et al. (2004) first prepared graphene by the scotch tape stripping method and won the 2010 Nobel Prize in Physics. But the efficiency of this method is extremely limited to laboratory research (Yi and Shen, 2015). The advantage of mechanical exfoliation is that high-quality graphene can be obtained, but the mechanical exfoliation theory is more complicated than other stripping methods, and the stripping process is affected by many uncontrollable factors (e.g., the thickness of the graphene, and horizontal size).

Chen et al. (2012) used the mechanical peeling method in the experimental study. In the N-methyl pyrrolidone (NMP) system, the shear force produced by high-speed mechanical exfoliation interacts with van der Waals force between graphite sheets in an inclined 45° test tube (Figure 3a), resulting in relative slippage between graphite sheets, good peeling effect of the dispersion system assisted by NMP and overall dispersion. The process does not introduce defects in graphene (Figures 3b–d). Paton et al. (2014) extended the theory of Chen et al. (2012) and proposed the theory of shear rate:

where γ_min is the shear rate, E_{S, G} and E_{S, L} are the surface energies of graphene and solvent, the surface energy of NMP is 69 mJm2, η is the viscosity of the liquid, and L is the length of the graphite sheet. L = 300–800 nm of graphene was measured by transmission electron microscopy (TEM), and the minimum shear rate γ_min ≈ 10⁴ s⁻¹ was calculated. E_{S, G} was calculated by Formula (1): E_{S, G} = 70.5–71 mJm2, which is very close to the surface potential energy (E_{S, G} = 69 mJm2) for the automatic exfoliation of graphene liquid phase. To verify the surface energy theory of graphene liquid-phase auto-exfoliation, many researchers have obtained the best surface energy through experiments, but the numerical values are different, and a unified theory is currently lacking (Kozbial et al., 2014). Single surface energy theory cannot completely explain the mechanism of graphene liquid-phase auto-exfoliation (Coleman, 2012) and thus has been replaced by the Hamaker constant theory.

Electrochemical exfoliation method has become one of the most researched potentials of graphene liquid-phase exfoliation theory (Parvez et al., 2014) because of its advantages of being fast and has high efficiency, environmental protection, and low cost, etc. (Parvez et al., 2014; Artur and Paolo, 2016; Li et al., 2016; Ping et al., 2017). Su et al. (2011) exfoliated graphite by electrochemical stripping to prepare graphene dispersion, characterized by AFM, TEM, scanning tunneling microscopy (STM), Raman and attenuated total reflection–Fourier-transform infrared spectroscopy (ATR-FTIR). More than 60% of graphene in the dispersion was AB stack. Bilayer graphene has a lateral dimension of 30 μm on the surface, and its electrical properties were superior to those of reduced graphene oxide (rGO). Shinde et al. (2016) used hydrodynamics to prepare graphene by the shear-assisted electrochemical exfoliation method for the first time in the study of the electrochemical exfoliation of graphene. The graphene prepared at a high potential was thick and contained cracks. At low voltage (<1 V) potential, shear force had a more obvious assistant effect on electrochemical exfoliation (the exfoliation mechanism is shown in Figure 4). The monolayer, bilayer, and trilayer in the dispersion graphene occupied a high proportion. The Raman characterization showed only a few defects on the surface of graphene (I_D/I_G = 0.21–0.32), and these defects were distributed at the edge of graphene. Shi et al. (2018) demonstrated a non-electrified electrochemical exfoliation method to produce high-quality graphene (NEEG) sheets. By direct electrochemical reaction between graphite powders and metallic Li in 1 M of LiPF₆/propylene carbonate electrolyte, continuous graphite exfoliation is achieved with a high yield of excess 80% without any consumption of electric energy. The as-prepared NEEG has high quality with few defects (I_D/I_G = 0.45), a high C/O ratio of 27.74, and excellent electronic conductivity of 102.5 S/cm. As conductive additives in Li₄Ti₅O₁₂ electrode for lithium ion batteries, NEEG contributes to higher capacity and initial coulombic efficiency than common thermally rGO. Therefore, electrochemical exfoliation method is promising for large-scale production of high-quality graphene nanosheets (Liu Y. J. et al., 2018).

In a review of the functionalization of non-covalent bonds of graphene, Georgakilas et al. (2016) reported the existence of π-π interactions: (1) π bonds exist between two molecules; (2) the interaction with two the geometry of the molecule is related; and (3) two molecules containing the π structure overlap, and the flatness of the surface of the nano-thin film of the two conjugated structures influences the π-π interaction. Wang et al. (2014) used the density functional theory of verified dispersion correction in combination with the zero-order symmetric adaptive perturbation theory to study the π-π interaction between graphene and aromatic molecules and to verify the dispersion-modified density functional theory. The zero-order symmetric adaptive perturbation theory of spin component scaling analyses the π-π interaction between graphene and aromatic molecules, which is believed to depend largely on the dispersion force and electrostatic interaction and is found to be more abundant. When a negative electron fluorine atom is substituted to an aromatic group, the π-π interaction becomes stronger and the liquid-phase dispersion property of graphene improves. Unfortunately, Wang et al. (2014) have only made a simple speculation on the π-π interaction and have not been able to further study it. In the field of liquid-phase dispersion of graphene, the nature of the interaction between π-π remains controversial to date (Ding et al., 2018). Wang X. D. et al. (2018) used strong non-covalent bonding (e.g., π-π interaction, hydrogen bonding, and GH–C interaction) between molecules to prevent the agglomeration of rGO lamellae and obtained uniformly dispersed rGO/poly(vinyl alcohol) (PVA) nanocomposites, as shown in Figure 5. The mechanical properties (including Young's modulus and tensile strength) and thermal properties of the composites can be significantly improved by adding a small amount of rGO into PVA.

Only some researchers speculated a reason for the formation of electron-rich exchange between the π-conjugated structure of graphene surface and aromatic molecules. In addition, the π-π interaction between aromatic molecules and graphene not only improves the liquid-phase dispersion but also adjusts the band gap to precisely control its electronic properties (Nduwimana and Wang, 2009; Zhang et al., 2011; Deka and Chowdhury, 2016). The π-π interaction method includes pyrene derivatives, biomolecules, CNT, and GO.

Pyrene derivatives were first used in the liquid-phase dispersion of CNTs. Most researchers used NMP as solvent and pyrene derivatives as dispersant to prepare CNTs (Ehli et al., 2006; Meuer et al., 2009) by ultrasound degradation. However, the liquid-phase dispersion theory for dispersing CNT on curved surfaces is not necessarily applicable to the liquid-phase dispersion of graphene on planar surfaces (Abo-Zahhad et al., 2016). O'Neill et al. (2011) used aminoethyl pyrene (PyMeNH₂) and 1,3,6,8-pyrene tetrasulfonate [Py(SO₃)₄] as dispersants to stabilize the dispersion of monolayer graphene in aqueous solution and prepared transparent conductive films. Parviz et al. (2012) used various pyrene derivatives to disperse graphene in a water system. The graphene dispersed in the pyrene derivative system had a dispersion stability of several months. Among different pyrene derivative systems, the effect of the sodium pyrene sulfonate (Py-SASS) dispersion system was the best. As shown in Figure 6A, the concentration of graphene in liquid medium was as high as 0.8–1 mg/ml, as shown in Table 1. Parviz et al. (2012) found that dispersants were automatically adsorbed onto the surface of graphene when the polarity of solvents and dispersants was significantly different. A study on the role of pyrene derivative functional groups in liquid-phase dispersion found that functional groups with high electronegativity effectively promote the adsorption of dispersants onto the surface of graphene (Figure 6B). Zhang (2016) synthesized a new dispersant, sodium pyrene-butane-1-sulfonate (Py-C₄-SASS), for the stripping and dispersion of graphene in aqueous media. Parviz et al. (2012) found through a comparative analysis that the new dispersant and Py-SASS, a bismuth-based graphene dispersant, could stably diffuse graphene in an aqueous medium. The exfoliation effect of the two dispersant systems was characterized by Raman and AFM spectroscopic methods (Figure 7A). The ratio of single layer and less layer ( ≤ 4 layers) was extremely high, and the radial dimension (0.5 μm) of Py-C₄-SASS-assisted exfoliated graphene is larger than the radial dimension (0.2 μm) of Py-SASS-assisted exfoliation. The UV-Vis spectroscopy was also performed to test the two dispersion systems. The concentration of graphene in the Py-C₄-SASS system was twice that of the Py-SASS system (Figure 7B). Unfortunately, Zhang failed to demonstrate the underlying cause of the difference in the ability to strip and disperse graphene between Py-SASS and Py-C₄-SASS. Chen C. et al. (2017) used poly(2-butylaniline) (P2BA) as a dispersant to peel off a few layers of graphene in organic solvents. The strong π-π interaction between P2BA with π conjugate structure and graphene solves the problem of graphene agglomeration in organic solvents.

Among the various reagents that interact with graphene, the advantages of biomolecules are low cost, environmental friendliness, and biocompatibility (Liu et al., 2017). Biomolecules that can interact with graphene include proteins and nucleotides (Ehli et al., 2006; Zhang et al., 2011; Deka and Chowdhury, 2016) (DNA/RNA), polysaccharides, and bile salts (Chang et al., 2010; Paredes and Villar-Rodil, 2016). The non-covalent bonding of protein and graphene in an aqueous medium produces a strong steric hindrance effect, which sharply reduces the free energy of graphene in the liquid medium, preventing the mutual graphene layers from each other. The rate of proximity is important to achieve uniform and stable dispersion of the graphene liquid phase (Bourlinos et al., 2009; Laaksonen et al., 2010). Flavin mononucleotides can also be strongly adsorbed onto the surface of rGO (Ayán-Varela et al., 2015; Yoon et al., 2015; Munuera et al., 2016). For example, Vashist and Luong (2015) studied target proteins and found that nucleotides DNA/RNA were highly adsorbed onto the surface of electrochemically and thermochemically reduced rGO. Compared with the surfactant dispersion system, although the concentration of graphene was much lower in the biomolecular system than in the surface active system (Seo et al., 2011; Notley, 2012; Zhu et al., 2014), the advantages of the nucleotide (DNA/RNA) dispersion system, such as biocompatibility and low cost, are not comparable with those of other dispersion systems (Georgakilas et al., 2012; Ayan-Varela et al., 2017).

As a 1D nanomaterial, CNTs form a six-membered ring structure from carbon atoms and contain plenty of conjugated large π bonds (Ehli et al., 2006; Li et al., 2018). It can interact with graphene through π-π interaction to form a steric hindrance effect (Zhang and Huang, 2000; Zhou and Zhang, 2015; Cao et al., 2017) to achieve the stable liquid-phase dispersion of graphene. Yue et al. (2014) prepared epoxy resin fillers by ultrasonic degradation. A study on the dispersion effects of graphene nanoplates (GNPs), CNTs, and GNP/CNT in liquid medium by the volume sedimentation method found that the GNP/CNT (2:8 mixture) system exhibits the best effect of graphene stable dispersion. As shown in Figure 8a, scanning electron microscopy (SEM) characterization showed that GNP and CNTs can prevent each other from agglomerating. Pure GNP system has higher dispersion than the GNP/CNT system, as shown in Figures 8b–g. Ji et al. (2016) made use of sulfonated CNT (SCNT) and successfully prepared a high concentration of graphene. SCNT played a tremendous role in not only suppressing the agglomeration of graphene but also bridging graphene sheets to improve the interlamellar conductivity. Compared with other dispersant systems in the liquid-phase dispersion of graphene, CNTs and graphene are excellent carbon nanomaterials, and their mixed dispersion system has a wide range of applications in energy storage and conduction (Cui and Li, 2002; Dimitrios et al., 2006; Tang et al., 2014; Zhao et al., 2014; Ji et al., 2016). In the process of preparing PS/PMMA composites, Chen J. W. et al. (2017) developed an effective thermodynamic method to balance the π-π interaction between PS and multi-walled CNTs and the PMMA between multi-walled CNTs. The dipole–dipole interaction with the carboxyl group enables uniform dispersion of the multi-walled CNTs in the composite. Fortunately, the mechanism of the π-π interaction can also be analyzed.

GO is amphiphilic (Geng et al., 2017) and reduces the solid–liquid interfacial energy in liquid media (Tian et al., 2011; Shao et al., 2014). Tung et al. (2016) used GO as a dispersant to prepare graphene by stripping graphite in liquid phase. GO as a surfactant not only assists graphite stripping but also renders it uniformly and stably dispersed in liquid phase through the π-π interaction. UV-Vis, SEM, and TEM analyses showed that the stripping yield of single-layer and less-layer graphene in the dispersion is high (40% of the initial graphite mass), and the quality of graphene is excellent (C/O = 21.5, IDIG = 0.12, conductivity 6.2 × 104ms).

The non-covalent bond with polymers can achieve both the dispersion of the graphene inorganic solvent and the dispersion of the organic solvent. For example, Choi et al. (2010) first prepared rGO dispersions in pure water by the chemical reduction of GO. Amino groups on polystyrene capped by polymer amine were successfully grafted with carboxylic acid groups remaining on the surface of rGO, forming non-covalent bonds with graphene and transforming graphene from the aqueous phase to the organic phase by ultrasonic degradation. After FTIR and Raman were analyzed, the protonated amino group of polystyrene interacts with the carboxylic acid group on the surface of rGO, realizing non-covalent bonding with graphene and improving the dispersion of graphene in organic media. Tchernook et al. (2014) used polyethylene nanoparticles to add a surfactant to the aqueous system to polymerize, and its dispersion process is shown in Figure 9. The test proves that the graphene is not modified in the aqueous polyethylene/graphene mixed system and the polyethylene nanoparticles disperse the graphene. It has a steric hindrance effect (Li et al., 2012) to achieve the uniform and stable dispersion of graphene. Chen J. et al. (2017) prepared water-borne graphene dispersions from graphite by electrochemical exfoliation with alkaline solution of hydrolyzed styrene–maleic anhydride copolymer as electrolyte. X-ray powder diffraction (XRD), AFM, and TEM analyses showed that monolayer graphene exists in graphene in the dispersions. FTIR and thermogravimetric analysis demonstrated that the hydrolyzed styrene–maleic anhydride polymer is combined with the graphene surface by a non-covalent bond to obtain a high concentration (>1 mg/ml) and stable aqueous graphene dispersion.

Surfactants can adjust the surface energy of graphene and different media to promote its stable dispersion. Surfactants have been widely studied as effective agents for the liquid-phase stable dispersion of graphene (Coleman, 2012; Zhao H. L. et al., 2018). Wei et al. (2011) compared the dispersing ability of different surfactants to graphene in the experiment of preparing graphene dispersion solution by exfoliation graphite in liquid phase. Results show that non-ionic surfactant PVP exerts greater dispersing ability, better biocompatibility, environmental friendliness, and lower cost than do other dispersants. Vadukumpully et al. (2011) applied the cationic surfactant hexadecyl trimethyl ammonium bromide (CTAB) as a dispersing agent to ultrasonically exfoliate graphite in a glacial acetic acid system, and they found that a large number of single-layer and less-layer graphene are stably dispersed in the N, N-dimethylformamide–CTAB system. Tkalya et al. (2012) reviewed the liquid-phase dispersion theory of graphene by surfactants and summarized the surface active agents, including CTAB, PVP, sodium dodecyl sulfate, sodium dodecylbenzenesulfonate (SDBS), lithium dodecyl sulfate, tetradecyl trimethyl ammonium bromide, sodium cholate, sodium deoxycholate, taurocholic acid, Tween 20, Tween 80, and Triton X-100, among others. Wang et al. (2017) used spray coating to prepare graphene films on glass and n-Si substrates through ultrasonic degradation and utilized anionic surfactant SDBS as a dispersant to prepare graphene dispersion in pure water. The liquid was subjected to UV-Vis analysis to obtain a stable graphene dispersion with a concentration of 15% SDBS. The surface morphology of the graphene has a blade-like edge structure. At the same time, the transmittance of visible light is more than 82% (about seven layers) (Selvakumar et al., 2018). This result proves that SDBS has strong exfoliation and stable dispersing ability to graphene.

The surfactant system exerts the best dispersion effect on graphene liquid-phase dispersion, but it cannot avoid introducing other side effects to the dispersion liquid system because of its own defects while solving graphene liquid-phase stable dispersion (De et al., 2010; Bepete et al., 2017; Shabafrooz et al., 2018). For example, Tkalya et al. (2012) found that non-ionic surfactants have long hydrophilic end and large volume characteristics. While preventing graphene agglomeration, the large volume effect of non-ionic surfactants can also lead to large distance between adjacent graphene sheets, which affects the conductivity of graphene and the permeability of graphene membranes in the dispersion system. Bepete et al. (2017) applied mechanical exfoliation theory to peel graphite after intercalation of the compound in water to obtain graphenide (negatively charged graphene) solutions. By using the advantages of outstanding volatility of tetrahydrofuran, graphene, and tetrahydrofuran can be mixed in water at room temperature to obtain a high concentration (0.16 g/L) and uniformly dispersed aqueous graphene dispersion. Bellani et al. (2019) prepared graphene dispersion with high concentration and excellent dispersion stability in NMP by wet-jet grinding exfoliation and solvent-exchange process. Compared with traditional exfoliation and dispersion technologies (e.g., ultrasonic exfoliation and mechanical peeling), the above process extremely improved the exfoliated efficiency of graphene and satisfied the industrial requirement of large-scale preparation graphene. Although the dispersion of graphene is not prepared by using any surfactant, the large scale makes use of tetrahydrofuran, and NMP in the industry poses a certain safety hazard to human beings (Rodriguez-Laguna et al., 2018; Shabafrooz et al., 2018).

Therefore, the future theory of choosing green dispersants or avoiding using dispersants is the focus of the study on graphene liquid dispersion. For example, Zhang et al. (2018) exfoliated and dispersed graphene with nanocrystalline cellulose (NCC). The addition of NCC considerably reduces the deposition of graphene in the dispersion system. The NCC dispersion method is an efficient, green, and scalable method for preparing graphene fractions. The method of dispersing liquid has a wide potential for the industrial application of graphene in the future. Wang S. F. et al. (2018) used Li⁺ and OH⁻ between graphene sheets. The positive charge of Li⁺ moves to the adjacent C atoms, and the adjacent OH⁻ interacts with it, forming a hydroxyl group added to the surface of the graphene and increasing the hydrophilicity of the graphene (Figure 10A). The graphene dispersion in an aqueous medium was improved without using a dispersant, and a graphene dispersion having a concentration of 0.09 mg/ml was prepared (Figures 10B,C).

Hydrogen bonding is a strong intermolecular force (bonding energy is 5–30 kJ/mol). The formation of intermolecular hydrogen bonds is profitable to the dispersion and dissolution between substances (Wang H. et al., 2018). Generally, the certain quantitative oxygen-containing groups will remain on the surface of graphene prepared by oxidation–reduction method (Aradhana et al., 2018) and ultrasonic graphite in water for a long time (Li et al., 2019). The oxygen-containing groups can produce strong hydrogen bond with dispersant–solvent system to improve the interfacial properties of graphene and achieve stable dispersion of graphene in liquid phase (Xu et al., 2017; Qin et al., 2018). For example, Yang X. M. et al. (2010) prepared a PVA/graphene mixture by dispersing rGO uniformly and stably in PVA. FTIR analysis was carried out because the residual oxygen-containing functional groups (such as carboxyl and hydroxyl) on the surface of rGO were combined with hydroxyl groups in PVA through hydrogen bonding, and the liquid-phase dispersion of graphene was well realized. León et al. (2013) used melamine as a dispersant to peel graphite in different liquid media (including organic solvents and aqueous solutions) by ball milling to obtain graphene dispersions with good dispersion stability. In the same solvent system, the exfoliation and dispersion ability of graphite with electron-rich benzene derivatives and triazine derivatives were tested, respectively. The present equivalent benzene derivatives do not exhibit the ability of peeling and dispersing graphite with melamine in the same solvent system. However, the aminotrigine system has a strong peeling and dispersing ability for graphene. The analysis and calculation showed that the aminotrigine derivatives could form a wide range of hydrogen bond 2D structures on the surface of graphene and has a good peeling and dispersing ability for graphene. The formation of hydrogen bonding makes it possible for the multi-point interaction of graphene surface in liquid medium. The stable combination of dispersant and oxygen-containing groups on graphene surface (as indicated by the green dotted line in Figure 5) can improve the dispersion stability of graphene in liquid medium.