Fabrication of multi-layer graphene by repeated transfer

Chemical vapor deposition (CVD) of graphene by using copper foil substrate^1,2 is being developed as a fundamental material supply technology, which is indispensable for the industrial use of graphene. Graphene has excellent characteristics in a wide range of physical properties, such as extremely high electron mobility, thermal conductivity, and mechanical strength, and is expected to be applied in various fields in combination with the ultimate thinness of the atomic layer. On the other hand, despite the attractive characteristics of graphene as described above, it is expected that there will be applications that are difficult to realize due to the insufficient thickness of the atomic layer. In such applications, there is a possibility of adaptation by forming multiple layers while maintaining the characteristics of single-layer graphene as much as possible. For that reason, the demand to control the number of graphene layers is increasing.

Graphene is expected as a material that overcomes the problem of high resistance of copper wiring due to the miniaturization of electronic devices. Single-layer graphene is known to have a mobility of 200 000 cm²/V s in the suspended state.^3,4 In this case, assuming a carrier density of 10¹²/cm², the electrical resistivity of 1 μΩ cm can be expected.⁵ This is a smaller value than silver, but in reality, the mobility is reduced due to the influences of lattice defects, grain boundaries, and silicon oxide substrates, and the expected conductivity is not obtained. Therefore, when using graphene for wiring, it is essential to laminate multiple layers to form parallel conduction and reduce resistance. In the case of the AB stacking of the graphite structure, a finite mass is generated, which deteriorates the mobility of carriers,^6,7 so the disoriented (turbostratic) stacking is an ideal multi-layer structure of graphene.⁸ 

In addition, since the single-layer graphene has no bandgap, it absorbs light over a very wide range of wavelengths from ultraviolet, visible, near-infrared, mid-infrared, far-infrared, and terahertz regions, and is expected as an optical sensor material. On the other hand, the light absorption rate of single-layer graphene is 2.3%. This value is remarkably large for the material of the atomic layer; however, it must be said that it is very small for the sensitivity of the optical sensor for practical uses. Several ideas have been proposed for improving the sensitivity of optical sensors by graphene.⁹ Multi-layer graphene formed with a disoriented structure, which maintains the band structure of single-layer graphene, is also an idea for improving the sensitivity.

Recently, it has been reported that multi-layered graphene can be formed by repeated transfer of single-layer graphene on a silicon oxide substrate and that it can exhibit higher mobility than single-layer graphene on a silicon oxide substrate.¹⁰ This is because the graphene in the second and third layers maintains the linear dispersion of the single-layer graphene due to the disoriented multi-layered structure. Furthermore, it was explained that the influence of non-uniform charge on the surface of the silicon oxide substrate is suppressed by graphene in the first layer so that the scattering of carriers in the second layer is suppressed and high mobility is exhibited. Multi-layer stacking by transfer is considered to be a reliable method for achieving the stacking of disoriented structures.

As mentioned above, multi-layer graphene is important to promote the industrial use of graphene. In this study, single-layer graphene was repeatedly transferred to a silicon oxide substrate to form multiple layers, and the doping characteristics associated with the multiple layers were observed. Furthermore, the structure of multi-layer graphene was investigated in comparison with pyrolytic graphite with a turbostratic structure.

A commercially available polycrystalline copper foil with a thickness of 10.5 µm was used as a catalyst substrate for single-layer graphene synthesis. The copper foil was set in a CVD apparatus equipped with an inductively coupled plasma source.¹¹ Copper foil was annealed at 600 °C in a hydrogen atmosphere in order to remove copper oxide on the surface and improve the flatness. Subsequently, methane gas was introduced into the CVD chamber, and single-layer graphene was grown on the copper foil by plasma irradiation at 1000 °C for 20 s.

The single-layer graphene was transferred from the copper foil to a commercial SiO₂/Si substrate (100-nm-thick oxide film/p⁺-silicon). The surface of SiO₂ was treated with UV ozone in order to remove hydrocarbon contamination and increase hydrophilicity. As a supporting polymer film for the transfer, a poly(methyl methacrylate) (PMMA) thin layer was spin-coated onto the synthesized graphene. The copper foil was etched with an ammonium persulfate aqueous solution, and the graphene with thin PMMA layer was sufficiently rinsed by replacing the etching solution with water while floating on the liquid surface. The graphene with thin PMMA layer on SiO₂/Si was dried, and then immersed in acetone and isopropyl alcohol to remove the PMMA coating. Careful removal of water during the drying process was performed to suppress the formation of graphene wrinkles.

In many studies, PMMA residues and residual water on the graphene surface after transfer were removed by annealing. Following this, in this study, graphene transferred to a silicon oxide substrate was annealed at 330 °C for 1 h in 100 Pa with 5% hydrogen and 95% argon using a quartz tube annealing furnace.

By repeating transfer and annealing, up to eight layers of multi-layer graphene were formed on the silicon oxide substrate. The crystalline quality of the graphene after every transfer and every annealing was evaluated using Raman spectroscopy with a laser wavelength of 532 nm for excitation and the spot size of 1 µm. To measure the interlayer distance of multi-layer graphene produced by transfer, x-ray diffraction (XRD) measurement using Cu-Kα was performed. Rigaku RINT2100XRD-DSII equipped with a horizontal goniometer (Rigaku Ultima III) was used.

Figure 1(a) shows a Raman spectrum of single-layer graphene transferred onto a silicon oxide substrate. The peak positions of the G band and 2D band were 1597 and 2686 cm⁻¹, respectively. Both peaks were fitted with single Lorenz curve. The peak widths (FWHM) of the G band and 2D band were 12 and 37 cm⁻¹, respectively. The Raman spectrum measured after the annealing of this graphene at 330 °C for 1 h is shown in Fig. 1(b). The peak positions of the G band and 2D band were 1594 and 2673 cm⁻¹, respectively. Both peaks were fitted with single Lorenz curve, and the peak widths of the G band and 2D band were 19 and 36 cm⁻¹, respectively. Intrinsic self-supporting single-layer graphene unaffected by strain and doping has been reported to have G band and 2D band wavenumbers of 1581.6 and 2676.9 cm⁻¹, respectively, for the excitation laser wavelength of 514.5 nm.¹² Considering that the 2D band shifts by 100 cm⁻¹/eV with respect to the energy of the incident laser, the wavenumber of the 2D band of the intrinsic self-supporting single-layer graphene excited by 532 nm used in this study is 2668.9 cm⁻¹.¹⁴ Compared with these wavenumbers of the G band and 2D band of the intrinsic self-supporting single-layer graphene, it can be seen that both the G band and 2D band are shifted to the higher frequencies after the transfer and the annealing. This suggests that the graphene transferred onto the silicon oxide substrate is significantly doped. In addition, the annealing hardly restored the G band and 2D band peaks of the Raman spectrum to the peak positions of intrinsic graphene.

Raman spectroscopy is a useful tool for the sensitive detection of charge doping and strain in graphene. Lee et al. reported a method for separating the charge doping and the strain in graphene by analyzing the correlation between the G band and 2D band in the Raman spectrum.¹² In this study, following their method, we analyzed the strain and the charge doping of graphene transferred onto the silicon oxide substrate. In Fig. 2, the wavenumber of the G band is taken on the horizontal axis and the wavenumber of the 2D band is taken on the vertical axis, and the correlation between the G band and 2D band of the Raman spectrum is shown. G-band and 2D-band wavenumbers of the Raman spectrum measured after each transfer and annealing were plotted. In Fig. 2, the red and black tilt axes are drawn. The red axis represents the mechanical distortion and the black represents the charge doping, which separates the wavenumber changes in the G band and 2D band. The G band and 2D band peaks move along the black axis due to the increase of generated charge due to p-type and n-type doping. Based on the wavenumbers of the G band (1581.6 cm⁻¹) and the 2D band (2676.9 cm⁻¹) of the intrinsic self-supporting single-layer graphene unaffected by strain and doping,¹² the origin is set to G = 1581.6 cm⁻¹ and 2D = 2668.9 cm⁻¹ by considering the shift of the 2D band by 100 cm⁻¹/eV with respect to the energy of the incident laser.¹³ The black “1” point in Fig. 2 is a plot of G band and 2D band wavenumbers of the Raman spectrum of graphene transferred to the silicon oxide substrate shown in Fig. 1(a). In this case, it can be seen that the charge of about 1 × 10¹³/cm² is doped. In addition, the compressive strain of 0.1%–0.2% remains in graphene. Annealing this graphene at 330 °C [the Raman spectrum of Fig. 1(b)] results in a red “1” point in Fig. 2. The doping situation changed little, which suggests that it is difficult to reduce the charge doping by annealing.

The black “2” in Fig. 2 is a plot of the wavenumbers of the G band and 2D band of the Raman spectrum measured after transferring the graphene of the second layer onto the graphene of the first layer. Graphene transferred to a silicon oxide substrate is generally doped in p-type, and there is concern that PMMA and water residue used for the transfer may have an influence. However, due to the second layer transferred using PMMA as in the first layer, the G band and 2D band of the Raman spectrum largely moved toward the origin in Fig. 2. Thus, the amount of doping was significantly reduced in the second layer of graphene. This suggests that the main cause of doping is the direct contact between the silicon oxide substrate and graphene, and the influence of PMMA and water residue used for transfer is small. It was reported that at the surface of thin silicon oxide grown on silicon, the oxide ions (SiO⁻) provide a layer of negative charges, which forms an electric double layer with the positive charge of sodium ions at a depth of about 20 nm from the surface.¹⁴ Furthermore, it is considered that the negative charges of the silanol group are distributed on the surface of the silicon oxide substrate by the UV ozone treatment process in order to remove the hydrocarbon contamination and to increase hydrophilicity.¹⁵ It is considered that a positive charge is induced in the graphene transferred onto silicon oxide by these negative charges. At the same time, this results in the screening of the negative charges on the silicon oxide substrate by the first layer of graphene.

By repeating the transfer and annealing of graphene in the third layer, the fourth layer, …, and the eighth layer, the plots of the wavenumbers of the G band and 2D band of the Raman spectra gathered in the upper left quadrant of the figure. This quadrant is a region where doping and mechanical strain cannot be discussed, but the wavenumber of the G band is close to the one of the intrinsic single-layer graphene.

To examine the characteristics of multi-layer graphene formed by repeated transfer, a comparison with pyrolytic graphite, which has a turbostratic structure (density 2.18–2.22 g/cm³, in-plane thermal conductivity 300 W/mK) was performed. Figure 3(a) shows a Raman spectrum of the pyrolytic graphite. The wavenumbers of the G band and 2D band are 1580 and 2699 cm⁻¹, respectively. The 2D band can be fitted by a single Lorenz curve with narrow width of 41 cm⁻¹ (FWHM), which is a feature of turbostratic graphite.¹⁶ Furthermore, the sharp D band that is also characteristic of turbostratic graphite is confirmed.¹⁶ When the wavenumbers of the G band and 2D band are plotted in Fig. 2, it is located close to the points of the multi-layer graphene formed by the repeated transfer. For reference, the Raman spectrum of a commercially available high-quality graphite sheet (density 2.10 g/cm³, thermal conductivity 1850 W/mK) is shown in Fig. 3(b). The 2D band is composed of two peaks, which is the characteristic of AB-stacked graphite.¹⁶ 

X-ray diffraction was used to measure the interlayer distance of multi-layer (four-layer) graphene by repeated transfer. Figures 4(a) and 4(b) show the x-ray diffraction spectra before and after the annealing of the multi-layer graphene, respectively. To suppress the background signal of x-ray diffraction, a single crystal sapphire substrate was used as a substrate for graphene stacking. The insets in Figs. 4(a) and 4(b) are the optical micrographs of four-layer graphene on sapphire substrates, which show good stacking with no noticeable tears or wrinkles. In both diffraction spectra, a wide peak can be confirmed centered around a 2θ angle of 25°. These peaks are attributed to the reflection of x rays between layers of multi-layer graphene formed by repeated transfer, which corresponds to the (002) reflection of graphite. The interlayer distances evaluated from the angle of these diffraction peaks were 0.350 nm for before the annealing and 0.348 nm for after the annealing.

Figure 5(a) shows an x-ray diffraction spectrum of a pyrolytic graphite that has a turbostratic structure. The peaks of (002) and (004) reflections are clearly confirmed, which indicates that this graphite has a highly oriented laminate structure. The interlayer distance evaluated from the angles of these (002) and (004) peaks was 0.342 nm. The graphitization rate (AB stacking rate¹⁷) evaluated from this interlayer distance was 17%, corresponding to the random stacking rate of 83%, which confirms the turbostratic layer structure. On the other hand, Fig. 5(b) shows an x-ray diffraction spectrum of a commercially available high-quality graphite sheet. The interlayer distance evaluated from the angles of peaks of (002) and (004) reflections was 0.336 nm, which corresponds to the AB stacking rate of 89%. As suggested by the two-peak structure of the 2D band of the Raman spectrum in Fig. 3(b), the x-ray diffraction measurements confirmed that this graphite sheet was predominantly AB-stacked.

Table I summarizes the interlayer distances and the random stacking rates for multi-layer (four-layer) graphene by repeated transfer, the pyrolytic graphite with turbostratic structure, and high-quality graphite sheets. It was reported that the interlayer distances of AB-stacked graphite and randomly oriented graphite are 0.335 and 0.344 nm, respectively.¹⁷ The interlayer distance of the high-quality graphite sheet was almost the same as the interlayer distance of AB-stacked graphite, and the interlayer distance of the pyrolytic graphite was almost the same as the interlayer distance of randomly oriented graphite. On the other hand, the interlayer distance of the multilayer (four-layer) graphene by repeated transfer was slightly larger than that of the randomly oriented graphite. This suggests that by repeating the transfer, the multi-layer graphene approaches the ideal randomly oriented graphite.

In this study, multi-layer graphene was formed by repeated transfer onto a silicon oxide substrate, and changes in its doping characteristics were observed. In addition, the structure of multi-layer graphene by repeated transfer was investigated in comparison with the pyrolytic graphite, which has a turbostratic structure. Single-layer graphene is doped due to the influence of the silicon oxide substrate, and the influence of PMMA and water residue, which are used for the transfer, is small. Graphene in the first layer suppresses the influence of the silicon oxide substrate, and the second and subsequent layers are almost unaffected. By repeating the transfer and stacking, multi-layer graphene approaches the ideal turbostratic graphite.

This work was supported by Innovative Science and Technology Initiative for Security Grant No. JPJ004596, ATLA, Japan.