Structural characterization of individual graphene sheets formed by arc discharge and their growth mechanisms

Abstract

Graphene sheets formed by arc discharge are hard to characterize in detail due to their complex pristine states in raw soot, which always exhibit an overall morphology of overlapping aggregation, together with other carbonaceous by-products. Here we used an improved arc method and simple separation procedure to obtain a large number of individual graphene sheets with single- to few-layers, and further probed their structural details using optical microscopy, transmission electron microscopy (TEM), atomic force microscopy and Raman spectroscopy. By TEM characterization, two major types of graphene sheets are shown; one is observed with folded fringes and polycrystalline structure, whereas the other is with an even graphene plane and single crystalline structure. In contrast to that of supported graphene, the Raman spectra of these graphene sheets show some different characteristics such as opposite shift of G band frequency as the layers increase. With increasing layers, the frequencies of G and G′ bands and the full width at half maximum (FWHM) of the G′ band totally exhibit layer-dependence. According to the FWHM of G′ bands, the folding within graphene sheets is also discussed. In addition, the defect types for arc graphene are analysed based on the D and D′ bands. Our results suggest that the D bands of such graphene sheets result from edges, rather than topological defects or disorder. Based on the findings, a new growth mechanism of arc graphene is proposed rationally responsible for the difference of two types of graphene sheets.

---

Introduction

The arc-discharge technique to vaporize graphitic carbon via a hot plasma opened a new branch of science and technology of nanocarbon materials. Fullerenes, especially C₆₀ and C₇₀, were first synthesized by a laser evaporation method in 1985, then for the sake of research and mass-production, the arc method was developed to prepare this material by Krätschmer et al.¹ Carbon nanotubes (CNTs) were discovered in arc-discharged samples and also developed to meet industrial demands using such arc method.² Graphene, a truly two-dimensional nanocarbon, has attracted intense interest due to its remarkable electronic, thermal, mechanical, and transparent properties.³ The prevailing synthetic methods used to produce graphene, such as mechanical exfoliation,⁴ epitaxial growth,⁵ liquid-phase exfoliation,⁶ and chemical vapor deposition (CVD),⁷ however, basically have inevitable drawbacks in the preparation process, transfer techniques, cost, quality or quantity. The arc method is a very promising way to produce high-crystallinity, doping-viable, few-defect graphene sheets via simple steps.^8–10 Synthesis by this method is usually carried out under a gaseous or liquid environment, creating a hot plasma to evaporate the anode material. The evaporated carbon radicals (mainly C₂ and C₃ in the plasma zone, whereas C₄, C₅… in areas distant from the discharge zone) coalesce and form carbon nanostructures depending on the reaction conditions.¹¹ The central temperature of the arc plasma can exceed 4000 K, which thermally anneals the products and removes topological defects during the growth process.

The overall morphology of few-layer graphene (FLG) fabricated by the arc plasma technique exhibits as a layer of overlapped sheets without any substrate. Most reported arc FLG sheets have some common structural characteristics as follows: a number of layers from 10 (10L) to bilayer (2L), lateral size ranging from 100 to 300 nm.^9,10,12–17 In the arcing process, the kinds of end-products are directly associated with the reactive conditions (buffer gas, catalyst, etc.), but the physical dimension and morphology are determined by the process of growth and deposition. People can manipulate arc plasma to control the growth process by adding an external magnetic field, resulting in a larger-area FLG (500–2500 nm) but with thicker layers of 3–15L.^18,19 The bottleneck to produce single-layer graphene (SLG) by arc discharge has been successfully solved by virtue of catalysts and rapid cooling.^13,20 The former provided 1–3L graphene sheets in the cathode deposit using SiC as carbon source and Cu as catalyst, whereas the latter produced fluffy soot containing 1–3L graphene collected from the inner wall of the vacuum chamber.

Arc graphene, and its doped derivatives^21,22 and hybrids (e.g. graphene/SnO₂,²³ graphene/CNTs²⁴) have been demonstrated experimentally to be very applicable for the fields of electronic devices, transparent conductive films, supercapacitors, and energy storage due to their high conductivity and specific surface area,^9,16 high-rate capability,²³ high capacity for hydrogen storage,²⁵ thermal stability^9,12 and excellent transparency.^13,16

Generally, previous works on the structure and properties of arc graphene are mainly based on aggregated graphene sheets either in raw soot or purified samples. To our knowledge, the structural details of individual graphene sheets, which act as basic units of this material, have barely been reported. However, we could not ignore the structural difference between the ensemble and the individuals. It is well-known that the electronic and band structures of two-dimensional graphene depend strongly on its lateral size, layer number, topological defects, edge types, folding and curvature, which in turn influence directly the properties of the material.²⁶ For example, arc graphene sheets with lateral size of 100–200 nm have better conductivity and higher rate capability than micro-scale graphene sheets.²³ In addition, we have noted that the experimentally measured properties of arc graphene fall within a wide range in the literature.^9,15,16,24 Regardless of synthetic parameters, the properties are probably associated with the structural details of the unit graphene sheets. Therefore, elucidating the detailed characteristics of this type of FLG is required not only for the overall understanding of the structure and properties but also for the development of arc-graphene-based applications. Moreover, the growth mechanism related to their basic units gives people guidelines and new insights into the nature of other nanostructures fabricated by arc plasma, more than nanocarbons.²⁷

To this end, we used an improved arc method and simple separation procedure to obtain a large number of individual graphene sheets from single- to few-layer, and further probed their structural characteristics with a combination of measurements of optical microscopy, transmission electron microscopy (TEM), high-resolution transmission electron microscopy (HRTEM), atomic force microscopy (AFM) and Raman spectroscopy. Our results show that two major types of graphene sheets with different configurations are formed simultaneously during the arcing process. By observation via TEM and HRTEM, folded fringes are shown to be a common feature for the majority of our samples, while the rest are basically ones with even surfaces. These folded fringes formed by curved or bent graphene planes are significantly different from those known for graphene sheets from other top-down methods, such as liquid-phase exfoliation.⁶ With increasing layer, arc graphene has the opposite frequency shift of the G band compared with supported graphene. Moreover, the frequencies of G and G′ bands, as well as the full width at half maximum (FWHM) of the G′ band totally exhibits layer-dependence. From Raman measurements, the defect types are discussed based on the D band intensity and corresponding intensity ratios. We demonstrated that the D band intensity of arc graphene results from edges, rather than topological defects or disorder. Based on these findings, we proposed a new growth mechanism responsible for the difference between the two types of graphene sheets formed in the arcing process.

Experimental

Briefly, an improved direct-current (DC) arc discharge system with an electric fan centrally facing the discharge zone was employed to carry out the preparation of raw FLG soot. Commercial pure graphite (99.99%) rods of 6 and 15 mm in diameter were set as anode and cathode, respectively. The electrodes were installed horizontally in a high-vacuum closed stainless steel chamber filled with a mixture of helium and hydrogen. Different sets of total pressures of He/H₂ buffer gases were used ranging from 20–90 kPa in our experiments. The optimized conditions for a high fraction of FLG are shown in Fig. S1 of ESI.† The direct current was maintained at 90–100 A and the discharge voltage was kept at 25–30 V by controlling the gap distance between the two electrodes. Raw soot samples were collected from the inner wall of the vacuum chamber.

The two-step separation procedures were carried out by an ultrasonic instrument (DR-MH20, Derui) and a high-speed centrifuge (TGL-16). The solvent contains a mixture of ethanol and deionized water in a volume ratio of 1 : 4. Due to the negatively charged Si substrate (hydrophilic), the aqueous solution can make the graphene sheets well dispersed.¹⁴ A solution of graphene with a concentration of 0.25 mg mL⁻¹ is processed by 2 h sonication and 30 min centrifugation with speeds of 4000, 8000 and 12 000 rpm. Then the upper part of the solution (supernatant) was transferred for the following measurements. Fig. S2† in ESI exhibits the TEM images of dispersed graphene sheets under three centrifugal speeds. With centrifugal speeds from 4000 to 12 000 rpm, the dispersion of graphene sheets can achieve a very high level to provide a large number of individual graphene sheets with varying layers.

The bright-field imaging and crystal diffraction of the samples were acquired by employing a TEM (JEOL-200CX) and a HRTEM (JEM-2010F). The acceleration voltages of TEM and HRTEM are 120 and 200 kV, respectively. The size of the selected-area aperture is ∼200 nm in our experiment. The surface morphology and thickness were probed by an AFM (Innova, Veeco) working at tapping mode on a silicon substrate. Raman spectra were recorded from 1000 to 3500 cm⁻¹ using Ar⁺ laser excitation wavelength of 514.5 nm (HR-800 laser confocal micro-Raman spectrometer, Horiba Jobin Yvon). The instrumental conditions are listed as follows: output: 20 mW, D1 filter; grating: 600 g × mm⁻¹; objective: ×100, and time: 10 s.

Results and discussion

1. TEM and AFM analyses

In Fig. 1a, one can see that arc graphene sheets are tangled with each other in raw soot. Fig. 1b illustrates the details of coexisted graphene sheets with varying layers via folded edges. As indicated by the arrows, single- (1L), tri- (3L) and multi-layer graphene sheets are shown clearly according to the folded edges of graphitic (002) planes. By the separation process mentioned in the experimental section, individual graphene sheets in sufficient quantity can be obtained from the raw soot. Fig. 1c and d exhibit the typical morphology of an individual graphene sheet and its selected area electron diffraction (SAED) pattern, respectively. Fig. 1c illustrates the morphology of such an individual graphene sheet where a part of the sheet has folded onto itself. According to the figure, the folded parts of graphene are wavy and turbostratic, indicating that the layer stacking is not commensurate and is disordered, possibly because of the uncontrolled growth process and the presence of other non-six-member carbon rings. The detailed growth mechanism will be discussed below. We found that the diffraction spots are weak and tend to be rings, indicating the polycrystallinity of this graphene sheet. The innermost spots correspond to a lattice spacing of 0.37 nm in the real space, larger than the 0.334 nm in graphite, denoting the changes of the graphene lattice. As shown in Fig. 1e, another typical graphene sheet has a relatively even surface with clear stacked edges as marked by the arrows. Fig. 1f exhibits the SAED pattern of the graphene sheet as depicted in Fig. 1e. The bright and sharp spots indicate the sixfold symmetry of six-member carbon rings in the basal plane, as well as the perfect lattice of this graphene sheet. Each set of diffraction spots in the same site is different in intensity, which indicates that the present graphene sheet is formed by the stacking of two component layers with different thickness.²⁸ The angle between the neighboring spots means these component layers stacked at a rotational degree of ∼6°. Diffraction spots with sixfold rotational symmetry show that electrons are incident normal to the basal plane of the crystal based on six-member carbon rings, allowing us to label the spots with Miller-Bravais indices (hkil). The innermost six spots correspond to the (10–10) set of lattice planes in the real space with a lattice spacing of 0.214 nm, and the outer spots correspond to the (11–20) set of planes with lattice spacing of 0.126 nm. Both lattice parameters are in agreement with the counterpart values of 0.213 and 0.123 nm of mechanically exfoliated SLG.²⁹ Fig. 1g is the AFM overview image of our sample within an area of 2.4 × 2.4 μm², showing good dispersion of these graphene sheets. The corresponding height profiles of L1 and L2 in Fig. 1g are shown in Fig. 1h, reflecting the height variation of the same graphene surfaces. The height curves indicated by green arrows mean graphene sheets with even surfaces, while the rest marked by red arrows denote those graphene sheets with uneven surfaces, further proving the presence of a folded structure. In addition, the mean thicknesses of 1L and 2L graphene sheets are about 0.5 and 0.9 nm, respectively.

Fig. 1 (a) Arc graphene in raw soot. (b) Magnified image of the box in (a), showing a jumble of single- (1L), tri- (3L) and multi-layer graphene sheets. (c) Typical morphology of an individual graphene sheet with folded fringes. (d) SAED pattern of (c), showing the polycrystalline structure. (e) Typical morphology of an even 2L graphene sheet with clear stacked edges as marked by the arrows. (f) SAED pattern of (e), showing sixfold symmetry and perfect lattice of the graphitic structure. The scale bars in (c) and (e) are 200 nm. (g) AFM image and (h) height profile show the height changes of graphene sheets marked by L1 and L2. Green and red arrows denote even and folded graphene surfaces, respectively.

As seen in Fig. 2a to d, HRTEM images illustrate the folded edges of 1L to 4L graphene sheets, respectively. In Fig. 2a and c, very even surfaces of graphene planes (SLG and trilayer) indicate the high quality and perfection of graphene sheets fabricated by arc plasma. In contrast, some of these graphene sheets are shown as bent and uneven surfaces of basal planes, indicating the formation of some topological defects during the growth process, as shown in Fig. 2b and d. In addition, a special structure is also observed (Fig. 2e), which resembles the overlapped crease in a multilayer graphene system, including pure and skewed stacking configurations.³⁰ Fig. 2f plots the relationship between interlayer spacing, d, and layer number, n of arc graphene. Obviously, a relaxation of van der Waals’s interaction between two neighboring layers (n ≤ 10) occurs since the interlayer spacing of 10L to 2L graphene increases from 0.34 to 0.41 nm. Hence only graphene sheets with layer number less than 10L can be defined as “few-layer” graphene.

Fig. 2 (a)–(d) HRTEM images of folded edges of 1L to 4L graphene sheets, showing the presence of even and bent graphene basal planes within the samples. (e) A crease-like structure in a 3-layer graphene system with an angle of 30°. (f) Plot of interlayer spacing, d₍₀₀₂₎, vs. layer number, n, indicating the relaxation of the interlayer bonding force. The line indicates the d₍₀₀₂₎ value of 0.334 nm for bulk graphite.

2. Raman spectra of individual graphene sheets

In the last decade, extensive research on Raman scattering of graphene has firmly established that the lineshape and the FWHM of G′ band, as well as the intensity ratio of I_G′/I_G, are unambiguously indicative of the layer number of graphene, especially for mechanically exfoliated graphene.³¹ In previous work, a large number of mechanically exfoliated AB-stacked graphene samples with micron sizes show substantial and distinguishable ranges of FWHM of G′ bands for single-, bi- and tri-layer graphene at 27.5 ± 3.8 cm⁻¹, 51.7 ± 1.7 cm⁻¹, and 56.2 ± 1.6 cm⁻¹, respectively.³² As shown in Fig. S3 of ESI,† however, arc graphene sheets, whose lateral sizes are much smaller than that of mechanically exfoliated graphene, even smaller than the laser spot size (∼1 μm) in Raman measurement, exhibit different but also straightforward features for determining the layer number. According to the contrast of optical transparency and corresponding FWHM of the G′ band, we can evaluate the layer number accurately. In our studies, graphene with a single-layer is almost transparent in optical observation and its lineshape symmetry, the I_G′/I_G ratio (∼1.2) and FWHM of G′ band (∼42 cm⁻¹) are consistent with that of liquid-phase exfoliated SLG sheet (∼300 nm in size, I_G′/I_G ∼ 1.5, FWHM ∼ 42 cm⁻¹).³³ Hence this is a solid evidence to demonstrate the accuracy of such assessment for layer number used in this work.

The optical microscopy image and typical Raman spectra of individual graphene sheets with layers of 1–3L are shown in Fig. 3a and b, respectively. In particular, the values of I_D/I_G ratio, I_G′/I_G ratio and FWHM of G′ bands are also listed accordingly in Fig. 3b. At 514.5 nm excitation, three major bands of G, D and G′ are always shown in the Raman spectra of graphitic and graphene materials. The G band (∼1582 cm⁻¹ for graphite) is a prominent feature for graphitic materials, which corresponds to the high-frequency E_2g phonon at the center (Γ point) of the Brillouin zone (BZ).³⁴ That is, the G band is related to the in-plane vibration of the graphene basal plane whose crystalline quality can be assessed by the FWHM of G band. The origin of the D band (∼1350 cm⁻¹) is related to the phonons of transverse optical (TO) branches around the corner of K point in BZ, activated by an intervalley double resonance (DR) process and dispersed strongly with excitation energy due to a Kohn anomaly at K.³⁵ The D band is associated with the A_1g breathing modes of six-member carbon rings, which requires defects or disorder for its activation.³⁴ Another prominent feature appearing at approximately 2700 cm⁻¹ for 514.5 nm laser excitation is called the G′ band, which is also known as the 2D band in the literature. The second-order G′ band in SLG arises from a DR process involving intervalley scattering by two iTO phonons.³⁶ This DR process gives rise to an intense G′ band which is larger than the G band in intensity.

Fig. 3 (a) Optical microscopy image of well-dispersed graphene sheets. (b) Typical Raman spectra of individual 1–3L graphene sheets, raw graphene soot and starting graphite excited by 514.5 nm laser excitation. The corresponding values of I_D/I_G, I_G′/I_G and FWHM of G′ bands are also listed. (c)–(e) show the average values of G band frequency (ω_G), G′ band frequency (ω_G′), and the FWHM of G′ band as a function of the inverse of layer number (1/n), respectively. The black solid hexagons indicate the values of bulk graphite.

In this work, the G band frequency, ω_G, downshifts by approximately 5 cm⁻¹ in the SLG sheet, i.e. ω_G(1) ∼ 1577 cm⁻¹ (see in Fig. 3b). That is, ω_G increases with increasing layer number, showing a slight variation in frequency (Δω_G ∼ 5 cm⁻¹). Thus the near linear layer-dependence can be presented as ω_G(n) = ω_G(graphite) + β/n, where β ≈ −4.78 cm⁻¹ is a constant. Interestingly, we found that the frequencies of G bands in our samples are shown to have an opposite shift (blue-shift) compared to supported graphene films, whose ω_G decreases (red-shift) as layer number increases.³⁷ As mentioned above, Raman scattering of the G band is related to in-plane vibrations, thus the shift of ω_G is most likely due to the changes of relative motions of in-plane carbon atoms. Doping, charged impurities or strain can lead to a blue or red shift of G band due to a change of the C–C distance.^37–39 The Δω_G values in the cases of doped charges and strain can reach a maximum of ∼9 cm⁻¹ and ∼5–6 cm⁻¹, respectively.^38,39 Because of the universality of the ω_G shift for arc graphene, as well as the nature of arc plasma, we inferred that the ω_G shift results from the external charges, which extensively exist within the arc plasma.^40,41 As the layers increase, the absorption of external charges is eliminated because of the recovery of the normal stacking order of graphitic layers.

In Fig. 3b, the D′ band split from the shoulder of the G peak is caused by an intravalley DR process, which is associated with graphite or graphene edges, or reduction in grain size.^42–44 As for individual 1–3L graphene sheets, the D′ bands are always obvious and easy to identify. With increasing layer, the D′ band gradually merges into the G band. In the raw soot sample and starting graphite, the D′ bands almost disappear due to the aggregation of samples. In Fig. S4 of ESI,† a weak peak appearing at ∼1450 cm⁻¹ is observed between D and G bands in the Raman spectra of some graphene sheets by accident. There are two explanations for the presence of this band in the literature; one is third order signals of the silicon substrate,⁴³ while the other is due to defect scattering.³⁷ These so-called D-scattering D₂ and D₃ bands are also observed in supported n-graphene layer films.

In contrast to the narrower FWHM of the G′ band for mechanically exfoliated SLG (∼24 cm⁻¹), we note that the FWHM for arc SLG is wider (∼41.5 cm⁻¹), indicating the broadening of the G′ band, which also occurs in other graphene sheets with similar size.³³ This is owing to symmetry lowering, which relaxes the selection rules for Raman scattering. Importantly, the information from the G′ band is sensitive to layer number because its Raman mechanism is closely related to the electronic band structure, which is changed with layer number and relative orientation.³¹ In our samples, with increasing layer, the frequency of the G′ band increases (blue-shift) and the FWHM becomes broader (see in Fig. 3c and d). The specific values of FWHM of G′ bands for 1–3L graphene are 41.2 ± 1.4 cm⁻¹, 46.4 ± 1.6 cm⁻¹, and 55.1 ± 1.5 cm⁻¹, respectively. Both the trends of frequency and FWHM for G′ band are consistent with the results in the literature.³⁶

The overtones of the 2D′ band centred at ∼3240 cm⁻¹ originates from a process with momentum conservation, which is satisfied by two phonons with opposite wavevectors. Both G′ and 2D′ bands always exist regardless of defects.³¹ Additional weak bands in the high-order region (ω ∼ 2450 and ∼2950 cm⁻¹) are assigned as D + D′′ and D + D′, respectively. D′′ derives from a phonon belonging to the longitudinal acoustics (LA) branch, seen at ∼1100 cm⁻¹ for visible excitation in defected samples.³⁴ This mode is probably associated with contributions from the region near the K point of the BZ boundary, and becomes Raman active due to the selection-rule relaxation resulting from defects. In addition, the D + D′ band is the combination of phonons with different momenta, around K and Γ, also requires a defect for its activation.³⁴ In Fig. 3b, it is clear that the 2D′, D + D′′ and D + D′ bands are relatively stable in both position and intensity.

As for folded and even graphene sheets, three typical Raman spectra of SLG, bilayer graphene and folded SLG are shown in Fig. 4a. The G band of the bilayer sample becomes more intense than even and folded SLG due to more carbon atoms contributing to this vibration mode under laser irradiation. The determination of folding within graphene also depends upon the symmetry and FWHM of G′ band. Due to arbitrary stacking, the G′ band of folded SLG exhibits a slight position shift, lineshape asymmetry and larger FWHM (∼50 cm⁻¹) than that of bilayer graphene (∼46 cm⁻¹), but smaller than that of trilayer (∼55 cm⁻¹) (see the inset of Fig. 4a). For even SLG, the G′ band can be fitted into only one Lorentzian peak (Fig. 4b), which represents the single π electron valence band and π* conduction band structure, and thus only one DR scattering cycle is excited near the K and K′ points (Fig. 4c). A whole DR cycle involves the four processes as follows: (i) laser-induced excitation of electron (e⁻)—hole (h⁺) pair; (ii) scattering (the first resonance process), and (iii) back-scattering of the excited electron (the second resonance process) by two phonons with opposite wavevectors q and −q; (iv) the recombination of the electron–hole pair. As shown in Fig. 4d, the G′ band for bilayer sample is dispersive and can be fitted into four Lorentzian peaks. In the bilayer system, the interaction between two layers causes the π and π* bands to divide into four parabolic band structures as π₁, π₂, π₁* and π₂*. Therefore, four Raman scattering cycles occur and their DR processes are denoted by the lines with two arrowheads in different colors in Fig. 4e. Each line represents a whole DR process. For multilayer system (≥3L), the electronic bands split into more complex and dispersive configurations, thus excited electron–hole pairs are involved in more scattering cycles, which contribute to a broader G′ band. In folded SLG, the G′ band has a similar frequency position to that of even SLG, but downshifts 4 cm⁻¹ compared to that of the bilayer sample. The change of G′ position for the folded part is also opposite compared with that of micron-sized graphene by mechanical exfoliation.³² That indicates the folded SLG exhibit a similar electronic structure to SLG but with increased Fermi velocity, which causes a larger slope of electronic band near K (K′) points, as illustrated in Fig. 4g. Previous studies show that misoriented bilayer graphene exhibits electronic structures that are similar to those of SLG but with reduced Fermi velocity, which means a smaller slope of dispersion curve near the K or K′ points.⁴⁵ With regard to the opposite frequency shift, we still consider linear π electron bands to explain its Raman scattering mechanism, as shown in Fig. 4g. The phonon with smaller wavevector q′ couples the Raman process that corresponds to a reduced slope of electronic sub-bands (dash lines), and thus lower frequency phonon dispersion near K points contributes to the G′ band.

Fig. 4 (a) Typical Raman spectra of SLG, bilayer graphene and folded SLG. (b) The single-Lorentzian fitted lineshape of G′ band of SLG, and (c) its DR mechanism of Raman scattering. (d) The fitted four components of the G′ band of bilayer graphene, and (e) the corresponding DR mechanism. (f) The fitted four components of the G′ band of folded SLG, and (g) the corresponding DR mechanism.

The intensity ratio of D/G bands, I_D/I_G, is widely used for characterizing the nanographite size L_a in graphitic materials. An empirical relation proposed by Tuinstra and Koenig (T–K) has been used as calibration for the qualitative control of structural transformation of graphitic materials.⁴⁶ The inverse of L_a is defined as defect density, n_D, which is used to denote quantitatively the degree of initial defects. Prior to discussing the intensity ratio I_D/I_G of our sample, L_a and L_D, which are frequently encountered in the literature, should be carefully distinguished. Basically, the measure of L_a depends on the amount of disorder in a nanocrystallite given by the amount of one-dimensional border with respect to total crystallite area. While the latter, L_D, denotes the distance between defects in graphene with zero-dimensional point defects. Therefore, L_a is suitable for line defects or polycrystalline graphene with domains, whereas L_D is more appropriate to describe the density of point-like defects within graphene. The correlations between L_D(n_D) and I_D/I_G are depicted as follows:⁴⁷

where E_L is laser energy, and equals 2.41 eV for 514.5 nm laser wavelength. Note that eqn (1) and (2) are only valid for graphene samples with point defects separated by L_D ≥ 10 nm using excitation light in the visible range.⁴⁷

According to HRTEM observations, the contributions to defect-induced D mode are hard to distinguish because the possible defects within this material include edge-like defects, topological defects (point defect, line defect), and fold-induced defects. Especially, folding may introduce more complex structural changes such as interlayer misorientation, and the bending or curvature of basal planes, which totally influence the Raman scattering process. To simplify the scenarios, the roles of each defect type in the contribution to D mode are discussed below separately. First, point defect is assumed to be the sole defect type in SLG, and then the L_D is in the range from 13.7 to 18.8 nm according to eqn (1). Compared to the order of magnitude of these graphene sheets in size (∼260 nm), the value of L_D indicates very low density of point defects and its contribution to D mode could be negligible. Similarly, without consideration of Raman relaxation length for point defect, a line defect can be seen topologically as a combination of some point defects, hence the low density of point defect will not lead to the massive presence of line defects. Second, as for fold-induced defects, Raman signatures caused by strongly folding (superlattice) will generate an additional I band at ∼1358 cm⁻¹ with respect to the D band,⁴⁸ or even a new R′ band centered at ∼1625 cm⁻¹.⁴⁹ Moreover, the fold-induced enhancement of D band intensity is only shown in the SLG system, i.e., between perfect SLG and folded SLG regions. In the FLG system, however, the impact of folding on the D mode is negligible.⁵⁰ These results strongly imply that the fold within arc graphene contributes little to the D band intensity even though it leads to a broader FWHM of the G′ band. Third, due to the massive presence of graphene edges, we inferred that the edges may play important role in the contribution to D bands. To validate this thought, three groups of measurements on the absolute intensities of D bands of graphene sheets with different numbers (N) are designed, as shown in Fig. 5a. Specifically, the cases as displayed in red (i), blue (ii) and green (iii) indicate the measurements on individual (N = 1), two (N = 2), three (N = 3) graphene sheets under laser irradiation, respectively. The laser spot size is jointly determined by excitation wavelength, λ, and effective numerical aperture, N.A. In this work, the laser spot size is about ∼1 μm. Fig. 5b shows the Raman spectra of the as-mentioned cases (i, ii, iii) and a reference Raman spectroscopy (black) measured on non-separated sample (N > 20) which occupies the whole area of the laser spot. For N = 1 samples, that is, individual graphene sheets, including 1L, 2L and 3L, totally show similar D intensities but increased G intensities due to increasing carbon atoms. In this case, the measured samples show near constant D band intensities (∼143 ± 35 cnt) but varying G band intensities, revealing no obvious variance of D band intensity with regard to layer number for individual graphene sheets. Moreover, the stable D intensity probably indicates that the fold-induced structural changes have little contribution to D band scattering. In the scenario of case (ii) (N = 2), adding one graphene sheet results in nearly twice larger intensity of the D band than that of case (i), equalling about 275 ± 30 cnt. Similarly, the D band intensity of three graphene sheets (N = 3) is approximately 381 ± 28 cnt. For non-separated sample (N > 20), the D band intensity achieve to above 600 cnt, much of which is probably due to the stacking disorder, rather than edges. As expected, the changes of D band intensities are shown as a near linear relationship with respect to increasing sheet number, N (the inset of Fig. 5b). Therefore, an edge-like defect is rationally considered as a dominant factor for D mode for arc graphene sheets.

Fig. 5 (a) Schematic of Raman measurements on individual to multiple graphene sheets and the corresponding optical microscopy image (scale bar: 1 μm). (i), (ii) and (iii) denote the cases of one (N = 1), two (N = 2) and three (N = 3) graphene sheets, respectively. (b) Typical Raman spectra exhibit the absolute intensities of D bands of individual to multiple graphene sheets. The red, blue and green curves were taken according to the cases of (i), (ii) and (iii) in (a), respectively. The black curve corresponds to the case of multiple graphene sheets (N > 20) which occupy the whole area of the laser spot. The inset in (b) gives the relationship between the mean absolute intensities of D bands and the sheet number (N), showing edge-dependent changes of D-band intensities.

On the other hand, the formation of defects and their types depend crucially on the process of crystal growth. For example, boundaries are always observed in CVD-grown polycrystalline graphene because simultaneous nucleation at different locations leads to independent domains, then a boundary appears when two neighbouring domains coalesce. In the arcing process, the high temperature (>4000 K) in the plasma zone facilitates the relaxation toward thermal equilibrium, and defects can anneal rapidly. Because of high formation and migration energies of point defects, it is unlikely that there are any new vacancies or adatoms after growth.⁵¹

In addition, the D′ band is also indicative of defects in graphene. As documented, the intensity ratio of D and D′ bands, I_D/I_D′, is strongly associated with the nature of the defect on the graphene surface.⁵² The I_D/I_D′ ratio depends on the type of defects, with edge-like or boundary defects being characterized by I_D/I_D′ ∼ 3.5, whereas point-like defects (vacancy) in the graphene basal plane give rise to I_D/I_D′ ∼ 7 and sp³-hybridized defects to ∼13.⁵² As shown in Fig. 6, the intensity ratio of I_D/I_G as a function of I_D′/I_G indicates that the values of I_D/I_D′ of 1L to multi-layer graphene sheets basically fall within the range between edge and point defects. From FLG to graphite, basal plane defects or disorder contribute largely to D mode, rather than edges. Moreover, the mean I_D/I_D′ value of 1L (∼4) to 2L (∼4.1) graphene sheets strongly suggest a very low density of point defects in the basal plane of such FLG sheets. We also noted that some of I_D/I_D′ data for 3L and multilayer graphene are distributed in the vicinity of vacancy line (∼7), indicating the possible presence of multi-type defects. Moreover, the Raman scattering processes of multilayer graphene, including 3L, are more complex than that of 1L or 2L graphene. Hence the I_D/I_D′ indicator may not be very accurate for thicker graphene. As for 1L or 2L graphene, their I_D/I_D′ data fall within the range of edge-like defect, which are in good agreement with the above discussion.

Fig. 6 The intensity ratio of I_D/I_G as a function of I_D′/I_G, showing the defect types of 1L and 2L fall within the range of edge defects. For 3L or multilayer graphene (>3L), some of I_D/I_D′ data are distributed in the vicinity of the vacancy line (∼7), indicating the possible presence of multi-type defects within this material.

Mechanically exfoliated graphene is always used as an ideal sample to investigate the intrinsic properties of graphene due to its defect-free lattice and Bernal stacking. Actually, the edge or boundary should be taken into consideration for most graphene samples such as CVD-grown graphene, graphene sheets and nanoribbons. For arc graphene sheets, small lateral size leads to the massive presence of edges within the material, thus the properties are probably different from that of large-area supported graphene. As shown in Fig. S5 of ESI,† graphene sheets fabricated by the techniques of liquid-phase exfoliation,^33,53–56 reduction of graphene oxide^57,58 and arc discharge,^{10,12–17,19} always exhibit lower I_G′/I_G ratios, broader FWHM of G′ bands, and the presence of the D band in contrast to mechanically exfoliated SLG.³¹ This indicates the electronic and vibrational properties of such graphene sheets have been changed by finite lateral size and massive edges. Moreover, the structural quality of arc graphene is comparable with that of liquid-phase exfoliated graphene, and superior to that of reduced graphene oxide.

3. Growth mechanism of arc graphene

High temperature and arc force are two major factors in the growth of carbon nanomaterials. High temperature will lead to the evaporation of the graphitic anode, resulting in carbon clusters with various atomicities. The arc forces usually consist of electromagnetic force (denoted as P_r in Fig. 7a), spot pressure and plasma fluid force.^40,41 Specifically, the former two forces are related to the ionization of gas molecules, as well as the evaporation and exfoliation of the anode, whereas the last one supplies gas molecules into the arc zone and pushes the ionized and evaporated matters into the ambient space, as shown in Fig. 7b. It is well known that the formation of CNTs and fullerene by arc discharge is crucially dependent on the presence of non-six-member carbon rings, which are indispensable for the curling/rolling of graphene planes.¹¹ On the other hand, the production of arc graphene on the chamber wall is widely believed to be due to the direct exfoliation of graphitic layers from the carbon anode.^9,12,13 In this work, however, we observed that two major types of graphene sheets coexist within the soot collected from the chamber wall, indicating a different mechanism compared with the previously proposed mechanism of arc graphene growth. That is, graphene sheets on the chamber wall result from the direct exfoliation of the graphitic anode due to the arc or gas pressure.¹² As shown in Fig. 7c, we inferred that two synthetic routes are probable for the graphene growth in the arcing process: (i) non-six-member carbon rings, including carbon pentagon (C₅) and heptagon (C₇), bond the exfoliated graphene (EG) together due to the presence of dangling bonds and introduce curvature. The dashed circle illustrates the “hinge” effect of non-six-member carbon rings. Hence, the products grown in this way show poly-crystallinity as observed by TEM. (ii) EG sheets are formed directly from the exfoliation of the graphitic anode, keeping the integrity of their graphene lattice in the whole process, which is also demonstrated by the TEM observation.

Fig. 7 Schematic of growth mechanism of arc graphene with two different structures. (a) Side and cross-sectional views of the carbon anode when generating arc force. I and P_r denote the electric current and electromagnetic force, respectively. (b) Recombination of carbon clusters with varying atomicities. Red, green and blue circles denote the pentagon, hexagon and heptagon of carbon atoms, respectively. EG means directly exfoliated graphene from the anode. (c) Two synthetic routes to graphene sheets in the arcing process: (i) non-six-member carbon rings bond EG together. The dashed circle illustrates the “hinge” effect of non-six-member carbon rings. (ii) The rest of the EG sheets directly deposit onto the chamber wall.

Conclusions

Raw soot containing a high-percentage of FLG sheets is synthesized using an improved arc discharge under optimized conditions of rapid-flow and buffer gas. Simple separation procedures enable such raw soot to eliminate most carbonaceous by-products and be well dispersed in graphene suspension. TEM and HRTEM observations show that two major types of graphene sheets coexist within the samples; one features folded fringes and a polycrystalline structure, whereas the other is has an even graphene plane and single crystalline structure. The folding configurations of graphene surfaces are also probed by AFM measurements, which demonstrate the presence of even and folded graphene sheets with varying heights. In Raman measurements, both the frequency and FWHM of the G′ band exhibits good layer-dependence, which is in agreement with that for supported graphene. The G frequency, however, is observed as having the opposite red-shift from graphite (1582 cm⁻¹) to SLG sheet (1577 cm⁻¹) in contrast to supported graphene films. The G-frequency shift probably results from the presence of excessive charges induced by the arc plasma. In addition, the G′ band of the folded SLG sheet shows a broader FWHM than that of bilayer graphene, but similar position to that of even SLG, indicating that their different Raman scattering mechanisms relate to DR models. Moreover, we demonstrated experimentally the D band of individual graphene mainly results from edge-like defects, rather than topological defects or disorder. This is evidenced by incremental absolute D intensities as the amount of graphene sheets increases under identical experimental conditions. The results of I_D/I_D′ analysis also imply that edge-like defects are dominant for 1–3L graphene sheets. The comparison of graphene sheets by different methods, i.e. arc graphene, liquid-phase exfoliated graphene and reduced graphene oxide, shows the high quality of our samples. Based on the findings, a new growth mechanism of arc graphene is proposed to be rationally responsible for the difference of the two types of graphene sheets. According to the nature of the arc plasma and the presence of non-six-member carbon rings, two synthetic routes occur simultaneously during the arcing process. One type of single crystal graphene with even surface is exfoliated directly from the graphite anode by arc force, whereas the other type of graphene with folded and polycrystalline structure is the combination of some small sheets linked by non-six-member carbon rings in the arc growth zone. In a word, our studies show that the structure of arc graphene is inhomogeneous in detail even though the samples are synthesized under identical conditions, which in turn leads to process control of arc plasma synthesis.

Acknowledgements

This work was financially supported by the Natural Science Foundation of Shaanxi Province and the Doctoral Research Assistant Foundation of Xi’an Jiaotong University. B. Li was personally supported by the Fundamental Research Funds for the Central Universities.

Notes and references

1. W. Krätschmer, L. D. Lamb, K. Fostiropoulos and D. R. Huffman, Nature, 1990, 347, 354–358.
2. Y. J. Su and Y. F. Zhang, Carbon, 2015, 83, 90–99.
3. K. S. Novoselov, V. I. Fal’ko, L. Colombo, P. R. Gellert, M. G. Schwab and K. Kim, Nature, 2012, 490, 192–200.
4. K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva and A. A. Firsov, Science, 2004, 306, 666–669.
5. C. Berger, Z. Song, X. Li and X. Wu, Science, 2006, 312, 1191–1196.
6. Y. Hernandez, V. Nicolosi, M. Lotya, F. M. Blighe, Z. Y. Sun, S. De, I. T. Mcgovern, B. Holland, M. Byrne, Y. K. Gun’ko, J. J. Boland, P. Niraj, G. S. Duesberg, S. Krishnamurthy, R. Goodhue, J. A. Hutchison, V. Scardaci, A. C. Ferrari and J. N. Coleman, Nat. Nanotechnol., 2008, 3, 563–568.
7. X. Li, W. Cai, J. An, S. Kim and J. Nah, Science, 2009, 324, 1312–1314.
8. N. Li, Z. Y. Wang, K. K. Zhao, Z. J. Shi, Z. N. Gu and S. K. Xu, Carbon, 2010, 48, 255–259.
9. Z. S. Wu, W. C. Ren, L. B. Gao, J. P. Zhao, Z. P. Chen, B. L. Liu, D. M. Tang, B. Yu, C. B. Jiang and H. M. Cheng, ACS Nano, 2009, 3, 411–417.
10. K. S. Subrahmanyam, L. S. Panchakarla, A. Govindaraj and C. N. R. Rao, J. Phys. Chem. C, 2009, 113, 4257–4259.
11. H. Lange, A. Huczko, M. Sioda and O. Louchev, J. Nanosci. Nanotechnol., 2003, 3, 51–62.
12. R. Kumar, R. K. Singh, P. K. Dubey, P. Kumar, R. S. Tiwari and I. K. Oh, J. Nanopart. Res., 2013, 15, 1847.
13. L. P. Huang, B. Wu, J. Y. Chen, Y. Z. Xue, D. C. Geng, Y. L. Guo, G. Yu and Y. Q. Liu, Small, 2013, 9, 1330–1335.
14. B. S. Shen, J. J. Ding, X. B. Yan, W. J. Feng, J. Li and Q. J. Xue, Appl. Surf. Sci., 2012, 258, 4523–4531.
15. Y. N. Chen, H. B. Zhao, L. M. Sheng, L. M. Yu, K. An, J. Q. Xu, Y. Ando and X. L. Zhao, Chem. Phys. Lett., 2012, 538, 72–76.
16. Y. P. Wu, B. Wang, Y. F. Ma, Y. Huang, N. Li, F. Zhang and Y. S. Chen, Nano Res., 2010, 3, 661–669.
17. Z. Y. Wang, N. Li, Z. J. Shi and Z. N. Gu, Nanotechnology, 2010, 21, 175602.
18. J. Li, A. Shashurin and M. Keidar, IEEE Trans. Plasma Sci., 2011, 39, 2366–2367.
19. I. Levchenko, O. Volotskova, A. Shashurin, Y. Raitses, K. Ostrikov and M. Keidar, Carbon, 2010, 48, 4570–4574.
20. B. Li, X. L. Song and P. Zhang, Carbon, 2014, 66, 426–435.
21. B. S. Shen, J. T. Chen, X. B. Yan and Q. J. Xue, RSC Adv., 2012, 2, 6761–6764.
22. S. A. Shah, L. Cui, K. Lin, T. Xue, Q. Guo, L. Li, L. Zhang, F. Zhang, F. Hu, X. Wang, H. Wang, X. Chen and S. Cui, RSC Adv., 2015, 5, 29230–29237.
23. Z. Y. Wang, H. Zhang, N. Li, Z. J. Shi, Z. N. Gu and G. P. Cao, Nano Res., 2010, 3, 748–756.
24. Y. P. Wu, T. F. Zhang, F. Zhang, Y. Wang, Y. F. Ma, Y. Huang, Y. Y. Liu and Y. S. Chen, Nano Energy, 2012, 1, 820–827.
25. G. F. Guo, H. Huang, F. H. Xue, C. J. Liu, H. T. Yu, X. Quan and X. L. Dong, Surf. Coat. Technol., 2013, 228, S120–S125.
26. P. T. Araujo, M. Terrones and M. S. Dresselhaus, Mater. Today, 2012, 15, 98–109.
27. J. Yu, J. Gao, F. H. Xue, X. Yu, H. Yu, X. Dong, H. Huang, A. Ding, X. Quan and G. Cao, RSC Adv., 2015, 5, 68714–68721.
28. S. Horiuchi, T. Gotou, M. Fujiwara, T. Asaka and T. Yokosawa, Appl. Phys. Lett., 2004, 84, 2403–2405.
29. J. C. Meyer, A. K. Geim, M. I. Katsnelson, K. S. Novoselov, D. Obergfell, C. Roth, C. Girit and A. Zettl, Solid State Commun., 2007, 143, 101–109.
30. A. W. Robertson, A. Bachmatiuk, Y. A. Wu, F. Schaffel, B. Buchner, M. H. Rummeli and J. H. Warner, ACS Nano, 2011, 5, 9984–9991.
31. A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth and A. K. Geim, Phys. Rev. Lett., 2006, 97, 187401.
32. Y. Hao, Y. Wang, L. Wang, Z. Ni, Z. Wang, R. Wang, C. K. Koo, Z. Shen and J. T. Thong, Small, 2010, 6, 195–200.
33. K. R. Paton, E. Varrla, C. Backes, R. J. Smith, U. Khan, A. O’Neill, C. Boland, M. Lotya, O. M. Istrate, P. King, T. Higgins, S. Barwich, P. May, P. Puczkarski, I. Ahmed, M. Moebius, H. Pettersson, E. Long, J. Coelho, S. E. O’Brien, E. K. McGuire, B. M. Sanchez, G. S. Duesberg, N. McEvoy, T. J. Pennycook, C. Downing, A. Crossley, V. Nicolosi and J. N. Coleman, Nat. Mater., 2014, 13, 624–630.
34. A. C. Ferrari and D. M. Basko, Nat. Nanotechnol., 2013, 8, 235–246.
35. S. Piscanec, M. Lazzeri, F. Mauri, A. C. Ferrari and J. Robertson, Phys. Rev. Lett., 2004, 93, 185503–185506.
36. L. M. Malard, M. A. Pimenta, G. Dresselhaus and M. S. Dresselhaus, Phys. Rep., 2009, 473, 51–87.
37. A. Gupta, G. Chen, P. Joshi, S. Tadigadapa and P. Eklund, Nano Lett., 2006, 6, 2667–2673.
38. C. Casiraghi, S. Pisana, K. S. Novoselov, A. K. Geim and A. C. Ferrari, Appl. Phys. Lett., 2007, 91, 233108.
39. E. del Corro, M. Taravillo and V. G. Baonza, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 85, 033407.
40. B. Gross, B. Grycz and K. Miklossy, Plasma technology, Iliffe Books Ltd, London, 1968.
41. D. A. Gerdeman and N. L. Hecht, Arc Plasma Technology in Materials Science, Springer-Verlag, Wien New York, 1972.
42. L. G. Cancado, M. A. Pimenta, B. R. Neves, M. S. Dantas and A. Jorio, Phys. Rev. Lett., 2004, 93, 247401–247404.
43. C. Casiraghi, A. Hartschuh, H. Qian, S. Piscanec, C. Georgi, A. Fasoli, K. S. Novoselov, D. M. Basko and A. C. Ferrari, Nano Lett., 2009, 9, 1433–1441.
44. L. G. Cancado, K. Takai, T. Enoki, M. Endo, Y. A. Kim, H. Mizusaki, A. Jorio, L. N. Coelho, R. Magalhaes-Paniago and M. A. Pimenta, Appl. Phys. Lett., 2006, 88, 163106.
45. J. M. Lopes dos Santos, N. M. Peres and A. H. Castro Neto, Phys. Rev. Lett., 2007, 99, 256802.
46. F. Tuinstra and J. L. Koenig, J. Chem. Phys., 1970, 53, 1126–1130.
47. L. G. Cancado, A. Jorio, E. Martins Ferreira, F. Stavale and C. A. Achete, Nano Lett., 2011, 11, 3190–3196.
48. A. Gupta, Y. Tang, V. H. Crespi and P. Eklund, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 82, 241406.
49. V. Carozo, C. M. Almeida, E. H. Ferreira, L. G. Cancado, C. A. Achete and A. Jorio, Nano Lett., 2011, 11, 4527–4534.
50. R. Podila, R. Rao, R. Tsuchikawa, M. Ishigami and A. M. Rao, ACS Nano, 2012, 6, 5784–5790.
51. F. Banhart, J. Kotakoski and A. V. Krasheninnikov, ACS Nano, 2011, 5, 26–41.
52. A. Eckmann, A. Felten, A. Mishchenko, L. Britnell, R. Krupke, K. S. Novoselov and C. Casiraghi, Nano Lett., 2012, 12, 3925–3930.
53. D. Nuvoli, L. Valentini, V. Alzari, S. Scognamillo, S. B. Bon, M. Piccinini, J. lllescas and A. Mariani, J. Mater. Chem., 2010, 21, 3428–3431.
54. J. Dong, B. Zeng, Y. Lan, S. Tian, Y. Shan and X. Liu, J. Nanosci. Nanotechnol., 2010, 10, 5051–5056.
55. W. Du, J. Lu, P. Sun, Y. Zhu and X. Jiang, Chem. Phys. Lett., 2013, 568–569, 198–201.
56. L. Zhu, X. Zhao, Y. Li, X. Yu, C. Li and Q. Zhang, Mater. Chem. Phys., 2013, 137, 984–990.
57. I. K. Moon, J. Lee, R. S. Ruoff and H. Lee, Nat. Commun., 2010, 1, 626.
58. M. Wojtoniszak, X. Chen, R. J. Kalenczuk, A. Wajda, J. Lapczuk and M. Kurzewski, Colloids Surf., B, 2012, 89, 79–85.

---

Footnote

† Electronic supplementary information (ESI) available: Detailed synthesis methods and parameters, TEM images of samples under varying centrifugal speeds, Raman spectra and comparison plot. See DOI: 10.1039/c5ra23990g

---