Letter pubs.acs.org/NanoLett
Remote Catalyzation for Direct Formation of Graphene Layers on Oxides Po-Yuan Teng,† Chun-Chieh Lu,† Kotone Akiyama-Hasegawa,‡ Yung-Chang Lin,† Chao-Hui Yeh,† Kazu Suenaga,‡ and Po-Wen Chiu*,† †
Department of Electrical Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8565, Japan
‡
S Supporting Information *
ABSTRACT: Direct deposition of high-quality graphene layers on insulating substrates such as SiO2 paves the way toward the development of graphene-based high-speed electronics. Here, we describe a novel growth technique that enables the direct deposition of graphene layers on SiO2 with crystalline quality potentially comparable to graphene grown on Cu foils using chemical vapor deposition (CVD). Rather than using Cu foils as substrates, our approach uses them to provide subliming Cu atoms in the CVD process. The prime feature of the proposed technique is remote catalyzation using floating Cu and H atoms for the decomposition of hydrocarbons. This allows for the direct graphitization of carbon radicals on oxide surfaces, forming isolated low-defect graphene layers without the need for postgrowth etching or evaporation of the metal catalyst. The defect density of the resulting graphene layers can be significantly reduced by tuning growth parameters such as the gas ratios, Cu surface areas, and substrateto-Cu distance. Under optimized conditions, graphene layers with nondiscernible Raman D peaks can be obtained when predeposited graphite flakes are used as seeds for extended growth. KEYWORDS: Graphene, metal free, chemical vapor deposition, floating catalyst
T
sacrificial Cu film, followed by evaporating Cu at an elevated temperature.14 The resulting graphene films were isolated on silicon substrates, but might be obscured by Cu residue. Ruemmeli et al. reported a direct synthesis of single- and multiple-layer nanographene on MgO substrates without the use of metal catalysts.15 Other insulating substrates such as SiN, Si, Al2O3, and SiO2 have also been reported as potential catalysts for graphitization.16−19 These graphene-like films catalyzed by oxides or nitrides generally suffer from incomplete decomposition of carbon feedstock that leads to an intense Raman D peak overlapped with G peak, indicative of low crystalline quality and of massive coexisting sp3 carbons. Here, we propose a novel approach to grow graphene on silicon oxides without predeposited metal catalysts such as Cu and Ni. The prime feature of the proposed approach is the offsite decomposition of carbon feedstock using floating Cu and H atoms as catalysts in a gas phase, followed by the graphitization of carbon atoms on oxide surfaces, thus allowing for the formation of high-quality graphene films potentially comparable with those grown on Cu foils. Figure 1 schematically illustrates
he roadmap for semiconductor devices envisages that carbon nanostructures, including carbon nanotubes and graphene, will emerge as key materials in postsilicon electronics. The exotic properties of graphene, especially its high carrier mobility (200 000 cm2/V·s)1,2 and current density (108 A/cm2)3 have prompted research interest into graphene’s application in semiconductor devices. There are two main areas where graphene is being considered in integrated circuits: as the channel material for field-effect transistors and as interconnects between transistors. Both applications prefer direct synthesis of graphene on silicon wafers without any transfer or etch-off process that might result in unintentional doping and degradation through the creation of wrinkles, cracks, or tears.4−6 Recently, chemical vapor deposition has been successfully developed for graphene growth on various transition metals.7−13 High-quality graphene can be easily obtained using low-pressure CVD, yielding predominantly single-layer or even single-domain graphene due to carbon’s low solubility in copper. This approach is highly compatible with current semiconductor manufacturing processes and provides a cost-effective route toward direct integration of graphene with existing silicon technologies. However, using the CVD approach to grow graphene on insulators is not so straightforward. Ismach et al. developed a one-step approach to bypass the wet transfer in which graphene was first grown on a © 2012 American Chemical Society
Received: November 15, 2011 Revised: January 4, 2012 Published: February 14, 2012 1379
dx.doi.org/10.1021/nl204024k | Nano Lett. 2012, 12, 1379−1384
Nano Letters
Letter
Figure 1. Schematic illustration of graphene growth mechanism involving decomposition of CH4 by floating Cu and H. Cu particles are subliming from the Cu foil at 1000 °C. Graphene starts growing on SiO2 substrates after obtaining a certain distance from the Cu foil.
the working mechanism of the growth. The floating catalysts were used to provide an alternative decomposition path that involves different transition states and lower activation energy, enabling reactions that would otherwise be blocked or slowed by a kinetic barrier in the direct catalyzation on oxide surfaces. The remote metal catalyzation results in the more complete decomposition of the hydrocarbons, thereby forming graphene with few defects and amorphous carbons. The proposed method depends on the fact that Cu is a metal that readily sublimes at the temperatures used to grow graphene in a typically CVD process. Providing Cu particles in sufficient quantity can bring about more successful collisions of reactants, thus raising the effectiveness of the decomposition of carbon feedstock in the gas phase. This resembles the growth of carbon nanotubes using ferrocene as a floating catalyst.20 To experimentally demonstrate this idea, we grew graphene in a horizontal tube furnace using atmospheric pressure CVD. Cu foils cleaned by acetic acid were placed in the upstream gas flow and heated up to 1000 °C. At this temperature, the equilibrium Cu vapor pressure is about 3 × 10−7 bar.21 Silicon substrates were placed downstream at least 3 cm away from the Cu foils. This distance is important and related to the graphene quality and growth rate. At ambient pressure, H2 (5 sccm), Ar (230 sccm), and CH4 (30 sccm) were supplied for 30 min growth during which carbon radicals or small fragments were generated and transported downstream, forming a continuous film on silicon substrates. Following growth, the samples were characterized by Raman spectroscopy, which reveals the doping, strain, and defect density of the grown films and also allows for the identification of single- and multiple-layer graphene. Other characterization methods include transmission electron microscopy (TEM), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), and transport techniques in a field-effect configuration. Figure 2a,b shows representative optical photographs of graphene grown directly on silica. These images show uniform color contrasts without any wrinkles or rippled structures. The uniform and flat coverage may imply a compressive strain on the graphene because a strong pinning is provided by the substrate22 in addition to the difference in thermal expansion coefficients for SiO2 (which contracts upon cooling) and graphene (which expands). Figure 2c shows the SEM image of a transferred graphene flake on silicon substrate. Wrinkles indicated by the arrows appear after the transfer. Also of note is the fact that unlike graphene grown on Cu foils, the graphene
Figure 2. (a,b) Direct deposition of graphene on fused quartz and SiO2 substrates. (c) Transferred graphene onto a SiO2 substrate. The arrows indicate graphene wrinkles. (d) Low-magnification TEM image of graphene. The dark spots correspond to SiOx nanoparticles. (e) Atomic structure of graphene lattice grown in the proposed approach. The TEM image was acquired from a graphene sheet using seeded growth. (f) Atomic structure of graphene lattice grown on Cu foils for comparison.
sheets easily break into small pieces in the polymer-assisted transfer process. This is attributed to the strong graphenesubstrate interaction that rips the graphene in the wet-etching process. This strong interaction also results in many SiOx nanoparticles being left behind on the graphene following the etching of the underlying substrate, as can be seen in the lowmagnification TEM image (Figure 2d). In the clean areas, a hexagonal carbon lattice with few defects shows the typical crystalline quality of the graphene grown in the proposed approach (Figure 2e). 1380
dx.doi.org/10.1021/nl204024k | Nano Lett. 2012, 12, 1379−1384
Nano Letters
Letter
Figure 3. Distance-dependent growth of CVD graphene on SiO2. (a) Growth setup. A small piece of Cu foil was used to supply the subliming Cu particles as a floating catalyst. The 14 substrates, each of which has the same dimensions, were lined up downstream of the gas flow. (b) Raman G peak intensity as a function of distance. The growth time was 30 min for SiO2 substrates. The lower panels show the full Raman spectra for some distances. (c) Light transmittance of graphene films as a function of wavelength for different film thicknesses obtained by controlling the growth time at a fixed sample-to-Cu distance. The sheet resistance is indicated for each curve. (d) Transmittance versus growth time curve for graphene layers grown on fused quartz substrates.
growth mechanism that might govern graphene growth on SiO2 is that the evaporated Cu particles are temporarily deposited on the SiO2 surface, catalyze the formation of graphene, and finally evaporate again gradually, leaving graphene only on the SiO2 substrate. This scenario can be verified by growing graphene at higher and lower temperatures with more and less evaporation of Cu. As shown in Figure S8 in the Supporting Information, no discernible 2D peaks are seen in the Raman spectra for T < 950 °C. This differentiates the growth mechanism of graphene on Cu, which enables the reaction in a temperature range between 850 and 1050 °C. With current growth method, a high-temperature process is essentially needed, implying that evaporated Cu plays a crucial role in forming graphene on SiO2. We further analyzed the Cu content on SiO2 surface in the different stages of the growth, which provides a detailed insight into the growth mechanism. As shown in the energy dispersive X-ray spectroscopy (EDX) and XPS spectra (Supporting Information, Figure S6 and S7), Cu was not found within the detectable resolution limit following short growth time (3.5 h). The triangular islands seem to orient in the same direction with respect to the gas flow, indicating that Cu particles participate actively in the CVD growth process (Supporting Information, Figures S2−S4). Test experiments were carried out to verify the role of Cu played in the growth (Supporting Information, Table S1). For example, replacing Cu (mp = 1084 °C) by Ni (mp = 1455 °C) or removing Cu in the growth results in neither graphene-like nor pyrolytic carbon films formed on SiO2. It should be noted that when the same experiments were carried out in an old quartz tube on which a Cu layer has been coated at the tube ends, clear graphitic signals can be readily seen in the Raman spectra. Another 1381
dx.doi.org/10.1021/nl204024k | Nano Lett. 2012, 12, 1379−1384
Nano Letters
Letter
wrinkles or humps. The corresponding Raman G map reveals the micrometer-scale inhomogeneity. The G intensity can be over an order of magnitude higher in some small areas, indicating the formation of multiple graphene layers due to incontrolled random nucleation on the bare oxide surface. In the corresponding D map (Figure 4d), we see a reduced D intensity compared to the films grown with a small-area Cu foil in Figure 3a. The D intensity differs from place to place and does not correspond with the G intensity map. The random nucleation and poor surface migration of carbon are presumably the major causes of defects in this growth setup. To further reduce graphene defects and to enhance the film uniformity, we added mechanically exfoliated graphite debris on SiO2 substrates in the setup displayed in Figure 5a. Raman
providing an easy measure of the film thickness in addition to the light transmittance (Figure 3c). There are no carbon signals in the first five samples. Weak G peaks start appearing from d > 3.0 cm. The IG then increases exponentially with d and reaches a peak at d = 5−6 cm after which the growth rate drops rapidly. The high D peak is the prominent feature of graphene layers grown by the setup shown in Figure 3a. The D to G peak intensity ratio, ID/IG, is found to be of little relevance to the sample-to-Cu distance. The defective graphene structure is attributed partly to the insufficient Cu in the current setup and partly to the uncontrolled random nucleation, the former of which causes the incomplete dehydrogenation of methane. It should be noted that graphene also grows on the Cu foil, which supplies floating catalysts. The graphene coverage does not impede the evaporation of Cu due to the penetrable feature of graphene.14 Figure 4a shows another setup to provide more Cu particles for dehydrogenation reactions. A 1 × 4 cm2 strip of Cu foil
Figure 5. Micro-Raman characterization of the CVD graphene grown on SiO2. (a) Growth setup. A strip of Cu foil surrounding along the tube wall was used to supply the subliming Cu particles as a floating catalyst. The substrate was placed 3 cm away from the Cu strip. Mechanically exfoliated graphite debris on the substrate was used as seeds for extended growth. (b) Optical image of the grown graphene with the area (20 × 20 μm2) marked for Raman mapping. Raman intensity maps of (c) G peak (1530−1650 cm−1), (d) D peak (1290− 1410 cm−1), and (e) 2D/G ratio. (f) Typical Raman spectrum of graphene grown in the vicinity of the graphite seed.
Figure 4. Micro-Raman characterization of the CVD graphene grown on SiO2. (a) Growth setup. A strip of Cu foil surrounding along the tube wall was used to supply the subliming Cu particles as a floating catalyst. The substrate was placed 3 cm away from the Cu strip. (b) An optical image of the grown graphene with the area (20 × 20 μm2) marked for Raman mapping. Raman intensity maps of (c) G peak (1530−1650 cm−1), (d) D peak (1290−1410 cm−1), and (e) 2D/G ratio. (f) The full Raman spectrum of the area marked in the Raman maps.
mappings in the vicinity of the predeposited graphite flakes are shown in Figure 5b−e. The few-layer graphite (2−3 nm) displays a relatively higher G intensity due to the multiple reflections and interference of light. In sharp contrast, other areas exhibit a low and uniform G intensity, indicative of graphene layers with uniform thickness. The most prominent feature of the resulting graphene is its very low defect density, as shown in the D map and the full spectrum therein. When the nucleation rate is low, the predeposited graphite can act as seeds for the extended growth of high-quality graphene, resembling the controlled growth of single-domain graphene.31 However, the D intensity may rise in areas further away from the seeds, similar to the case shown in Figure 4. A higher
surrounding the quartz tube was placed 3 cm ahead of the SiO2 substrate. To gain more insight into the crystalline quality of the graphene films, Raman mapping (using 488 nm excitation) was performed with a low power intensity of
