Impact of Amorphous-C/Ni Multilayers on Ni-Induced Layer Exchange for Multilayer Graphene on Insulators.

Layer exchange growth of amorphous carbon (a-C) is a unique technique for fabricating high-quality multilayer graphene (MLG) on insulators at low temperatures. We investigated the effects of the a-C/Ni multilayer structure on the quality of MLG formed by Ni-induced layer exchange. The crystal quality and electrical conductivity of MLG improved dramatically as the number of a-C/Ni multilayers increased. A 600 °C-annealed sample in which 15 layers of 4-nm-thick a-C and 0.5-nm-thick Ni were laminated recorded an electrical conductivity of 1430 S/cm. This value is close to that of highly oriented pyrolytic graphite synthesized at approximately 3000 °C. This improvement is likely related to the bond weakening in a-C due to the screening effect of Ni. We expect that these results will contribute to low-temperature synthesis of MLG using a solid-phase reaction with metals.

Introduction

The technology to synthesize graphene on arbitrary substrates is essential to combine advanced electronic devices and carbon materials.[1] Mechanical transfer has allowed for high-crystallinity graphene on insulators; however, concerns regarding imperfections in graphene, such as wrinkles, remain. This has motivated researchers to develop a transfer-free process for practical applications. Moreover, for devices which can endure large currents or heat, thick multilayer graphene (MLG) is preferable to few-layer graphene.[2−4] For example, low-resistance wiring requires a thickness of tens of nanometers to replace Cu wiring in large-scale integrated circuits, whereas a heat spreader requires a thickness of more than several micrometers, depending on the type of device. Vapor deposition techniques (e.g., chemical vapor deposition[5−10] and plasma-assisted vapor deposition[11,12]) are the most promising ways to grow MLG directly on insulators. However, forming thick, uniform MLG is difficult. Conversely, metal-induced solid-phase crystallization of amorphous carbon (a-C) or polymers has been actively studied for the direct synthesis of MLG on insulators.[13−27] Some of these techniques have enabled the synthesis of thick (>5 nm) MLG by controlling the initial thickness of a-C.[18−27] However, further investigations are required to achieve high-quality MLG on insulators.

Recently, we developed metal-induced layer exchange (MILE) of a-C.[28−32] In particular, MILE using Ni enabled us to synthesize uniform MLG at low temperature with a wide range of thicknesses.[29,32] Higashi et al. reported the effects of an a-Ge/Au multilayer structure on the layer exchange between Au and Ge.[33,34] The grain size of the resulting Ge layer was significantly improved by using the multilayer structure. The improvement was attributed to the diffusion of Ge atoms promoted by the bond weakening of a-Ge due to the screening effect.[35,36] In this study, we have applied the multilayer structure to the Ni-induced layer exchange of a-C, which has improved the crystal quality and electrical conductivity of the resulting MLG layer dramatically.

Results and Discussion

Figure a shows the schematic of the sample preparation procedure. A Ni thin film (60 nm thickness) was prepared on quartz glass (SiO2) substrates. Subsequently, a-C/Ni multilayers were prepared, where the number of a-C layers n was 1, 3, 5, and 15. The thickness of the Ni layers between the a-C layers were fixed at 0.5 nm. Each a-C layer thickness t was 60 nm for n = 1, 20 nm for n = 3, 12 nm for n = 5, and 4 nm for n = 15, as summarized in Table . Therefore, the total thickness of a-C layers was fixed at 60 nm.

Figure 1

(a) Schematic of the sample preparation procedure. (b) Scanning electron microscopy (SEM) image and (c) energy dispersive X-ray (EDX) spectrum of the sample for n = 15 nm and t = 4 annealed at 600 °C after Ni removal. The EDX spectrum was obtained in a field of view of 1 mm square.

Table 1

Relationship between the a-C Thickness t and the Number of a-C Layers na

t (nm)	60	20	12	4
n	1	3	5	15

Total thickness of the a-C layer is 60 nm.

Figure b shows that MLG covers almost all of the substrate, although there are some voids. Such void formation can be suppressed by increasing the film thickness ratio of a-C to Ni.[30]Figure c shows the peaks corresponding to C, Si, and O. These elements originate from the MLG layer and SiO2 substrate because the detection depth is greater than 1 μm in the energy dispersive X-ray (EDX) measurement. Note that the Ni concentration in the MLG layer is less than the EDX detection limit (∼1%), which will be low enough not to contribute to the electrical conductivity σ. Such low metal contamination in the resulting layer is a typical feature of MILE, which is limited by solid solubility.[31]

We investigated the detailed cross-sectional structure of the sample for n = 15 and t = 4 annealed at 600 °C. The cross-sectional transmission electron microscopy (TEM) sample was prepared by the conventional focused ion beam method. The bright-field TEM image in Figure a and the elemental map in Figure b show that a single Ni layer is uniformly stacked on a single C layer. We note that there is no remnant of the a-C/Ni multilayer prepared on the upper part before annealing. The selected-area electron diffraction (SAED) pattern in Figure c reveals that C{002} planes in MLG are oriented almost parallel to the SiO2 substrate. Compared to using a single a-C layer,[29] the current MLG has less ripple and higher quality. As shown in Figure d, the dark-field TEM image shows a bright contrast of the MLG layer in the entire region, which also indicates that MLG is highly (002) oriented. The high-resolution TEM image in Figure e shows that the lower part of MLG consists of small graphene layers, whereas Figure f shows that the upper part consists of relatively large graphene layers. The insertion of an AlO interlayer between the a-C/Ni multilayer and Ni will further improve the crystal quality of MLG.[30,32]

Figure 2

Characterization of the cross-section of the sample for n = 15 nm and t = 4 annealed at 600 °C before Ni removal. (a) Bright-field TEM image. (b) EDX elemental map. (c) SAED pattern taken from the region including the Ni and MLG layers with a selected area of 200 nm diameter. (d) Dark-field TEM image using the C{002} plane reflection. (e, f) High-resolution lattice images showing the (e) lower and (f) upper parts of the MLG layer, respectively.

Figure a shows that Raman spectra of the back side of the samples have peaks at approximately 1350, 1580, and 2700 cm–1, corresponding to the disordered mode (D), graphitic mode (G), and D mode overtone (2D) peaks in the graphitic structure, respectively.[37] This means that layer exchange between the C and Ni layers occurred, and MLG formed on the SiO2 substrate in all samples. The intensity ratio of the G to D peaks (IG/ID) in Raman spectra corresponds to the crystal quality of MLG.[37]Figure b shows that the IG/ID ratio increases as n increases for both annealing temperatures. The IG/ID ratio reaches the highest value of approximately 8 for the sample for n = 15 and t = 4 annealed at 600 °C. This result suggests that the crystal quality of MLG improves as n increases. According to previous studies on Au-induced layer exchange of Ge, the crystal quality improvement by the multilayer structure is attributed to the grain size enlargement due to the Ge diffusion enhancement.[33,34]Figure c indicates that σ also increases as n increases for both annealing temperatures. This behavior clearly reflects the crystal quality of the MLG found in the Raman study (Figure b). The 600 °C sample has a higher σ than the 800 °C sample for n ≥ 3. This behavior is likely attributed to the larger grain size due to the lower growth temperature, which is well known for layer exchange. σ exhibits the maximum value of 1430 S/cm in a sample, where n = 15 and t = 4, annealed at 600 °C. This value is close to that of highly oriented pyrolytic graphite (1700 S/cm) synthesized at approximately 3000 °C. According to our previous paper,[32] σ can be improved to 2700 S/cm by inserting an AlO interlayer and controlling the MLG thickness. By combining these techniques, σ will further improve.

Figure 3

Raman study and the electrical properties of MLG formed by layer exchange. (a) Raman spectra obtained from the back side of the samples prior to Ni removal. (b) IG/ID ratio of the samples determined by the Raman spectra shown in (a), and (c) electrical conductivity σ of MLG after Ni removal, as a function of n and t.

Conclusions

Adoption of the a-C/Ni multilayer structure greatly influenced the quality of MLG formed by layer exchange. The IG/ID ratio in Raman spectra and σ of MLG dramatically improved as the number of a-C/Ni multilayers increased. This behavior is likely because the higher number of multilayers provided the stronger screening effect and then the higher diffusion rate of C atoms into Ni, resulting in the larger grain size of MLG. Therefore, the present study is expected to contribute to low-temperature synthesis of MLG using solid-phase reaction with metals.

Experimental Section

Sample Preparation

All depositions were carried out using radio frequency (RF) magnetron sputtering (base pressure: 3.0 × 10–4 Pa) with Ar plasma. The substrate temperature was 200 °C for Ni and room temperature for a-C/Ni multilayer. The RF power was set to 100 W for a-C and 50 W for Ni. The samples were annealed at 600 °C for 2 h and 800 °C for 10 min in ambient Ar to form MLG on the substrate by layer exchange. To remove the Ni that moved to the top layer, the sample was dipped in a diluted HNO3 solution (2.0% HNO3) for 10 min.

Material Characterization

Scanning electron microscopy (SEM) analyses and EDX spectroscopy were performed using Hitachi High-Technologies SU-8020 and JEOL JEO-2300. TEM analyses were performed using an analytical TEM, FEI Tecnai Osiris operating at 200 kV, equipped with an EDX spectrometer (FEI Super-X system). Raman spectroscopy was performed using JASCO NRS-5100, wherein the laser wavelength was 532 nm and the spot size was 5 μm. The σ was evaluated by the van der Pauw method using a Bio-Rad HL5500PC system and averaged over five measurements for each sample.