1 © 2017 IOP Publishing Ltd Printed in the UK

1. Introduction

Over the last decade, graphene has gained enormous attention as a next generation electronic material, due to outstanding properties such as an exceptionally high carrier mobility (~250 000 cm2 (V · s)−1) and ballistic transport at room temper ature [1, 2]. Graphene also has excellent mechanical properties (Young’s modulus value of ~1 TPa), which makes

it a potential material for both micro- and nano-electrome-chanical systems (MEMS/NEMS) [3].

For the production of graphene-based electronic devices, high quality graphene on large area substrates with effective control of both structural and electrical properties is required. Chemical vapor deposition (CVD) growth on transition metals and dielectric surfaces [4, 5], and thermal decomposi-tion of bulk silicon carbide (SiC) substrates [6] were normally

Journal of Physics D: Applied Physics

Solid source growth of graphene with Ni–Cu catalysts: towards high quality in situ graphene on silicon

Neeraj Mishra1, John J Boeckl2, Anton Tadich3,4, Robert T Jones4, Paul J Pigram4, Mark Edmonds5, Michael S Fuhrer5, Barbara M Nichols6 and Francesca Iacopi1,7

1 Environmental Futures Research Institute, Griffith University, Nathan, 4111 QLD, Australia2 Materials and Manufacturing Directorate, Air Force Research Laboratories, Wright-Patterson AFB, 45433 OH, United States of America3 Australian Synchrotron, 800 Blackburn Road, Clayton, 3168 VIC, Australia4 Department of Chemistry and Physics, Centre for Materials and Surface Science, La Trobe University, 3086 Victoria, Australia5 Centre for Atomically Thin Materials and School of Physics and Astronomy, Monash University, Clayton, 3800 VIC, Australia6 US Army Research Laboratory, 2800 Powder Mill Road, Adelphi, 20783 MD, United States of America7 School of Computing and Communications, University of Technology Sydney, Broadway, 2007 NSW, Australia

E-mail: francesca.iacopi@uts.edu.au

Received 7 November 2016, revised 9 December 2016Accepted for publication 29 December 2016Published 1 February 2017

AbstractWe obtain a monolayer graphene on epitaxial silicon carbide on silicon substrates via solid source growth mediated by a thin Ni–Cu alloy. Raman spectroscopy consistently shows an ID/IG band ratio as low as ~0.2, indicating that the graphene obtained through this method is to-date the best quality monolayer grown on epitaxial silicon carbide films on silicon. We describe the key steps behind the graphene synthesis on the basis of extensive physical, chemical and morphological analyses. We conclude that (1) the oxidation, amorphisation and silicidation of the silicon carbide surface mediated by the Ni, (2) the liquid-phase epitaxial growth of graphene as well as (3) the self-limiting graphitization provided the molten Cu catalyst, are key characteristics of this novel synthesis method.

Keywords: graphene, 3C–SiC, silicon, liquid phase growth, soft x-rays

S Supplementary material for this article is available online

(Some figures may appear in colour only in the online journal)

N Mishra et al

Printed in the UK

095302

JPAPBE

© 2017 IOP Publishing Ltd

50

J. Phys. D: Appl. Phys.

JPD

10.1088/1361-6463/aa560b

Paper

9

Journal of Physics D: Applied Physics

IOP

2017

1361-6463

1361-6463/17/095302+9$33.00

doi:10.1088/1361-6463/aa560bJ. Phys. D: Appl. Phys. 50 (2017) 095302 (9pp)

N Mishra et al

2

used to obtain high quality, scalable graphene for industrial applications. Of both methods, CVD growth on metal sub-strates has been commonly used to produce large-area, high-quality monolayer graphene. However, the graphene needs to be transferred onto a semiconducting or insulating substrate which degrades the electrical and mechanical performance of devices. To address this issue, researchers have proposed direct growth of graphene on semiconducting surfaces such as germanium (Ge) layer on silicon (Si) wafers [7, 8]. Although a high-quality monolayer graphene is obtained in this case, the weak adhesion between the graphene and the underlying Ge layer is a serious drawback for the fabrication of functional devices.

On the other hand, growth via thermal decomposition of bulk silicon carbide (SiC) at high temperature (1300–1700 °C) provides the most promising pathway towards clean, structur-ally coherent and mechanically reliable epitaxial graphene at the wafer scale [6, 9, 10]. However, SiC substrates are very expensive, difficult to micro-machine and available in relatively small size (150 mm) as compared to silicon wafers (300–450 mm) [11, 12]. Growth of graphene on hetero- epitaxial cubic silicon carbide (3C–SiC) films grown on low-cost silicon (Si) substrates has been explored to address these issues [13–15]. Graphene growth on 3C–SiC/Si substrates offers compatibility with established Si fabrication technolo-gies, allowing for seamless integration with a large number of electronic and opto-electronic technologies. Although the thermal decomposition of 3C–SiC on Si to produce graphene appears promising for large-scale production of graphene-based electronic devices, it has been mostly limited to the use of 3C–SiC(1 1 1) oriented surfaces [15–17]. The quality of the graphene produced via this method is limited by the high defect densities of hetero-epitaxially grown SiC on silicon, and difficulty in controlling the Si sublimation rates in high/ultrahigh vacuum at relatively low (1100–1300 °C) temper-atures [16, 18].

In parallel, several research groups have investigated a catalyst-based method for obtaining graphene on the 3C–SiC surface [19–21]. This method involves depositing a thin nickel film on the SiC surface and subsequently annealing the sample at temperatures ranging from 750 °C to 1200 °C; a temperature much lower than that for the sublimation pro-cess. However, the quality of the graphene formed is often found to be strongly non-uniform. In most cases graphene was observed to grow on the metal surface and so required transfer to a semiconductor or an insulating surface to obtain a func-tional device. This transfer requirement hinders its utility for large-scale device fabrication [19, 21].

We have recently pioneered an alternative, alloy-mediated approach for direct and self-aligned synthesis of high quality graphene on 3C–SiC/Si [22, 23]. In this process, the carbon is supplied from the solid-phase using thin epitaxial SiC films on Si as the source. In contrast with the thermal decompo-sition based route, our alloy-mediated graphitization takes place at comparatively moderate temperatures (900–1100 °C) and moderate vacuum (~10−4 mbar). The graphene obtained exhibits a high adhesion energy with the underlying SiC sub-strate [22], estimated to be about an order of magnitude higher

than that of graphene transferred onto a SiO2 layer on Si [24]. Furthermore, this solid-source alloy-mediated approach ben-efits from straightforward wafer-level patterning capabilities and warrants a more detailed understanding of its synthesis mechanisms. It is anticipated that both the Ni–Cu alloy and the residual partial pressure of oxygen in the vacuum chamber play a crucial catalytic role, affecting the uniformity and quality of the graphene layer. Herein we focus on the mech-anism of alloy-mediated graphene growth on 3C–SiC on Si (1 0 0) substrates. We identify the oxidation and amorphization of the SiC as important intermediate steps of the solid-source graphitization, and the key role of the Cu catalyst component, responsible for an efficient and self-limiting process.

2. Experimental

2.1. Sample preparation

Hetero-epitaxial 3C–SiC films, 250 nm thick, were grown on 150 mm diameter Si(1 0 0) wafers in a hot-wall horizontal low pressure CVD reactor through alternate supply epitaxy com-prised of SiH4 and C3H6 at 1000 °C and a carbonization pro-cess at 950 °C [25]. The 3C–SiC/Si wafers were then diced into 10 mm × 10 mm samples for further experiments and characterization.

Following growth of the SiC, either a single layer of nickel or double layer of nickel and copper were sputtered on the 3C–SiC surface. For the nickel growth, Ni (99.99%) was deposited via Ar+ ion DC magnetron sputtering (Surrey Nano Systems) at 100 °C and a base pressure of 10−8 mbar with thicknesses of ~5, 8, 11 and 14 nm. For the double layer growth, an additional layer of ~15 nm copper, Cu (99.95%) was deposited on top of the Ni layer using a DC Ar+ ion sput-tering system (EMITECH K575X) with a deposition current of 100 mA at a base pressure of 10−4 mbar.

The alloy-mediated graphitization was performed in a high temperature, low vacuum tube furnace (Carbolite HT tube furnace), evacuated to a pressure below 5 × 10−4 mbar. The samples were heated and maintained at 1100 °C for 75 min and left to cool to room temperature under vacuum. A highly intermixed layer of unreacted metal, metal silicides and excess carbon is produced during the graphitization process, which was then selectively removed by immersing the samples in a wet etch solution (Freckle etch) [23] for 20 h, followed by rinsing with Milli-Q water. All subsequent characterizations were performed after the removal of this intermixed layer.

2.2. Sample characterization

In order to characterize the quality of the graphene grown on the sample surfaces, Raman spectroscopy was performed at room temperature in a backscattering geometry using a Renishaw InVia spectrometer operating at 514.5 nm (argon–ion laser) with a spot size of approximately 1 µm2. For calibration, we used a silicon reference sample producing a standard peak at ~520.5 cm−1 [26]. The D, G and 2D bands in the Raman spec-trum of the graphene layers were monitored and the intensity ratio of the D- and G-bands (ID/IG) was calculated. In order

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

3

to improve the statistical accuracy of the ID/IG ratio, Raman spectra were recorded at five different sites (at the centre as well as the four corners) on each 10 mm × 10 mm sample. In addition, confocal Raman mapping was performed using a WITec ALPHA300 RA CRM/AFM operating at 532 nm laser (1 mW power measured before the objective). The scan size, step size, number of points and integration time were 40 µm × 40 µm, 0.25 µm, 25 600 and 0.75 s respectively.

Atomic force microscopy (AFM) was performed in non-contact mode using a Park NX20 atomic force microscopy system over 5 × 5 and 1 × 1 µm2 scan areas with 512 × 512 pixel resolution.

High-resolution transmission electron microscopy (HRTEM) of the graphene/3C–SiC/Si sample was performed using an FEI Cs-corrected Titan3 microscope operating at 80 keV. Cross-sectional TEM specimens (sample foils) were prepared via a focused ion beam (FIB) liftout technique using a FEI Strata DB235 FIB/SEM with a Ga+ ion source. The foils were excavated from the bulk samples and thinned to about 500 nm. Subsequently, Ar+ ion milling was conducted in a Fiscione NanoMill™ to remove Ga+ ion damage. A 2 µm thick Pt/Au protective layer was deposited onto the samples prior to FIB milling.

The chemical composition and elemental depth profile of the graphene/3C–SiC/Si layered system was characterized using multiple techniques. Time-of-flight secondary-ion mass spectrometry (ToF-SIMS) was performed using an ION-ToF GmbH TOF.SIMS 5 instrument. Data was acquired from a 100 µm × 100 µm area using a 30 keV +Bi3 pulsed primary-ion beam, whilst sputtering a 300 µm × 300 µm area centered at the same location with 2 keV Cs+ ions. ToF-SIMS profile showed that nearly 650 s is required to etch the entire 250 nm SiC film. Therefore, the etch rate was calculated around ~23 nm min−1. The base pressure during the acquisition was 3 × 10−10 mbar.

X-ray photoelectron spectroscopy (XPS) was employed for depth profiling the elemental composition of the gra-phitized sample. The samples were initially sputter etched and characterized using a laboratory-based XPS instrument, Kratos AXIS Nova x-ray photoelectron spectrometer (Kratos Analytical Ltd) equipped with a monochromated Al Kα x-ray

source (hν = 1486.6 eV) operating at 150 W. Spectra were recorded over a binding-energy range of 0–1300 eV, at a pass energy of 160 eV, while sputtering a 3 mm × 3 mm area with 1 keV Ar+ ions (target current 0.1–0.2 µA). Spectral acquisi-tions were limited to a 110 µm × 110 µm area at the centre of the sputtered region. Prior to sputtering, the pressure in the analysis chamber was 2.6 × 10−9 mbar. The sputter-etch rate was calculated by measuring the actual crater depth using optical profilometry. The crater depth was measured to be ~110 nm, corresponds to a sputter rate of ~2.8 Å min−1.

Complementary synchrotron-based XPS measurements were performed at the soft x-ray beamline at the Australian Synchrotron. Samples were introduced into the ultra-high vacuum system at the beamline and annealed at 200 °C for several hours in order to remove airborne contaminants and ensure a pristine surface. Following this, the XPS measure-ments were performed using a SPECS Phoibos 150 hemi-spherical analyser, operating at a pass energy of 10 eV. Detailed C 1s and Si 2p core levels were measured at four different photon energies (150 eV, 350 eV, 850 eV, and finally 1486.6 eV); by tuning the photon energy in this way, the pho-toelectron mean free path within the material was systemati-cally changed allowing for variation in the sampling depth of the XPS measurement without the need for sputtering the sample.

3. Results

3.1. Raman spectroscopy

Figure 1(a) compares the Raman ID/IG band ratios (intensity ratio of the D- and G-bands) of the graphene grown on 3C–SiC films on Si (1 0 0) substrates using a Ni layer only, with those obtained using a Ni/Cu double layer. Note that the thick-ness of the Cu layer remained constant (~15 nm) when the Ni/Cu double layer is used. The error bar represents the standard deviation of the ID/IG ratios, measured at five different sites over the 10 mm × 10 mm graphene samples. Data obtained from samples produced over more than 10 experimental runs fell within the displayed error bars. The average ID/IG ratio of the graphene grown using only a Ni layer varies between

Figure 1. (a) Raman intensity ratio ID/IG comparison of graphene grown on 3C–SiC film using Ni only (red circles) and a Ni/Cu double layer (black squares), where the thickness of the Cu layer was kept constant at ~15 nm. The use of a Ni/Cu alloy leads to a substantial reduction in the ID/IG ratio, indicative of less defective graphene. (b) Raman spectra of a graphene layer obtained on the selected sample (table 1) showing D, G and 2D peaks at ~1355, 1584 and 2708 cm−1. (c) 2D band fitted using single Lorentzian peak with FWHM of ~40 ± 3 cm−1 indicate a monolayer graphene on top.

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

4

0.6 and 1. This ratio drops to about 0.2–0.3 when graphene is grown using the Ni and Cu layers. These results confirm our recent findings [22] that an additional Cu layer deposited on top of the Ni film significantly improves the graphene quality. Note that the ID/IG ratio shown in table 1 (0.24 ± 0.05) is con-siderably lower than our previously reported values (0.5–0.8) [22, 23], thanks to an improved control of the graphitization process.

We selected a sample prepared under the conditions which yielded the lowest ID/IG ratio (8 nm Ni/15 nm Cu) for further characterization. The ID/IG ratio along with the peak position and full width half maximum (FWHM) of the G and 2D peak of the corresponding sample are reported in table 1.

Figure 1(b) shows the Raman spectra of the selected sample, indicating the D, G and 2D bands at ~1355, 1585 and 2707 cm−1, respectively. The G band is located at ~1585 cm−1 with a FWHM of about ~25 cm−1. The position of G band is very similar to that for graphene on a bulk SiC surface (1586 cm−1; FWHM <30 cm−1), as reported in table 1. The intensity of the 2D band is slightly higher than the G band, indicating the presence of monolayer graphene [27].

Previous studies [28, 29] have reported that exfoliated monolayer graphene exhibits a relatively narrow and intense 2D band (FWHM <30 cm−1) at ~2678 cm−1. Figure  1(c) shows that the 2D band of our graphene sample is symmet-rical and fitted well using a single Lorentzian peak centered at 2707 cm−1 with a FWHM of ∼40 cm−1 (table 1), consistent with the characteristics of a monolayer graphene. However, in comparison to exfoliated graphene, the 2D band of our sample is significantly shifted towards higher frequency by approximately 30 cm−1. This blue shift could be related either to electron/hole doping or the strain induced by the substrate [28, 30]. As the dependence on doping of the 2D band shift is very weak [31], the significant blue shift of the 2D band of our samples is likely a consequence of a compressive thermal strain induced in the graphene upon cooling [28, 32]. Note that the Raman spectra acquired on the graphene/3C–SiC sample (figure 1(b)) is remarkably consistent over the entire 10 mm × 10 mm surface.

Additionally, a confocal Raman mapping was performed, measuring 25 600 points over a 40 µm × 40 µm area (see supplementary information, figure  S1 (stacks.iop.org/JPhysD/50/095302/mmedia)). The average ID/IG ratio, I2D/IG

ratio and 2D FWHM is calculated as ~0.4, ~1 and ~45 cm−1 respectively (table 1), the latter figures  both suggesting the prevalence of monolayer graphene distributed over the sur-face. Note that the intensity ratio of 2D to G band alone would be a weak indication of single layer as reported by Tiberj et al [36]. To confirm the number of graphene layers, the sample is further characterized with the help of HRTEM, ToF-SIMS and soft x-ray photoelectron spectroscopy. Although the ID/IG values calculated herein do not yet match those corresponding to the best quality CVD graphene or graphene grown on bulk SiC via thermal decomposition [5, 10], they represent the highest quality graphene grown on 3C–SiC/Si reported to date [14, 16, 34, 35]. For example, Fukidome et al reported average ID/IG values as low as ~0.4 for graphene grown on 3C–SiC(1 1 1)/Si(1 1 0), however, multilayer graphene (6–7 layers) was observed in their experiments [15]. In addi-tion, the crystallite size (La) calculated using the expression [37] ( ) ( ) ( / )λ= × − −L I Inm 2.4 10a

10laser4

D G1 (λ = 514.5 nm

in this work) indicates an average graphene domain size of around 35–60 nm, considerably larger than the crystallite size (La = 10–15 nm) reported for the graphene grown on 3C–SiC on Si via thermal decomposition method [16]. Therefore this catalytic alloy approach yields to-date the best quality, large area graphene grown in situ on 3C–SiC films on silicon wafers.

3.2. Transmission electron microscopy (TEM) and low energy electron diffraction (LEED)

Figure 2(a) shows the dark-field cross-sectional TEM micro-graph of the graphene/3C–SiC sample. The top dark layer is the protective gold and platinum material deposited prior to FIB preparation of the TEM cross sections. The middle lighter contrast region corresponds to a diffuse TEM diffrac-tion pattern (not shown here), consistent with an amorphous material approximately 20 nm thick. However, complemen-tary LEED measurements performed on this surface (figures 2(b) and (c)) reveal a clear square symmetry pattern with a high background. Since LEED is extremely surface sensitive (a few atomic layers at most) [38], this result indicates that regions of ordered crystalline symmetry exist at the surface (at least within the sampling depth of the LEED). These regions exhibit the same four-fold rotation symmetry as the 3C–SiC

Table 1. A comparison of the ID/IG ratio, peak position and FWHM (G and 2D band) of the graphene grown in this work against previously reported graphene growth methods: exfoliated, and grown via thermal decomposition on bulk SiC and 3C–SiC on Si.

Material Method Raman measurement ID/IG

Peak position (cm−1) FWHM (cm−1)

G 2D G 2D

Graphene/3C–SiC/Si(1 0 0) (this work)

Alloy-mediated (8 nm Ni/15 nm Cu) graphene growth

5 points on 1 × 1 cm2 0.24 ± 0.05 1585 ± 1 2707 ± 1 24 ± 3 40 ± 325 600 points on 40 × 40 µm2

0.40 ± 0.10 1588 ± 1 2703 ± 1 29 ± 1 45 ± 1

Exfoliated graphene [33]

Mechanical exfoliation <0.1 1580 2678 — <30

Graphene/SiC(0 0 0 1) [10]

Confined sublimation <0.1 ~1586 ~2701 — <30

Graphene/3C–SiC/Si [14, 16, 34, 35]

Thermal decomposition of 3C–SiC on Si in UHV

>0.4 1600 ± 10 2715 ± 5 60 ± 10 100 ± 10

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

5

surface. The TEM image also shows an atomic sheet resting above this ~20 nm thick amorphous layer, underneath the pro-tective layer. This sheet is most certainly the monolayer gra-phene, which is further confirmed by investigating chemical profile of the top few nm of the surface in following sections. Towards the bottom of the image, the single crystalline 3C–SiC(1 0 0) film is observed, characterized by the presence of typical stacking faults travelling along the 1 1 1 planes [25].

3.3. Time of flight secondary ion mass spectromertry (ToF-SIMS)

Changes in the chemical composition as a function of depth for the selected graphene/3C–SiC sample (table 1) were investigated with ToF-SIMS depth profiling (figure 3). The figure  plots the intensity (normalized to maximum) of the C+, Si+, SiC+ and Cs2O+ secondary ion mass fragments as a function of depth, from the surface, through the graphene layer, and into the 3C–SiC layer underneath. Note that Cs2O+ is formed by association of any oxygen atoms present in the sample with Cs+ ions from the sputter source and allows oxygen, which normally ionizes with a negative charge, to be profiled simultaneously with the positively charged C+, Si+, and SiC+ ions.

Since the sample surfaces can be contaminated with adven-titious hydrocarbons from atmospheric exposure, data from top ~1 nm is eliminated from the profile. A marked decrease in the C+ ions intensity over the next ~1 nm of the profile prob-ably attributed to the monolayer graphene at the surface. It appears from the Cs2O+ profile that there is extensive oxida-tion of the SiC layer down to a depth of ~20 nm. This thickness corresponds well with the amorphous region observed using

TEM between the graphene and the pristine 3C–SiC layers in figure 2(a). A relatively constant SiC+ signal is also obtained from the surface to a depth of around 30 nm. It is important to note that these ions can be generated from either pure SiC or oxidized SiC (oxycarbide), supplementary the hypothesis of the existence of crystalline SiC grains in an amorphous oxidised region, as observed in the TEM and LEED images (figure 2).

3.4. X-ray photoelectron spectroscopy (XPS) depth profiling

Seeking further insight into the nature of the amorphous region beneath the graphene, we have undertaken additional chemical depth profiling using x-ray photoelectron spectr-oscopy (XPS), performed using two XPS methodologies.

Figure 4 shows an ion-etched XPS depth profile taken from a laboratory-based XPS instrument (Kratos AXIS Nova x-ray photoelectron spectrometer), extending from the surface, through the graphene (indicated by the sharp peak in the C pro-file at the top surface) and into the 3C–SiC film. The elemental profiles reveal once again a significant amount of oxygen (>10 atomic %) present in the top ~20 nm of the sample along with carbon and silicon, indicative of an oxidized top region. This oxygen profile again corresponds roughly to the total thickness of the amorphous region observed with TEM in figure 2(a). In particular, we note that the top ~5 nm of the sur-face just underneath the graphene contains up to 35% oxygen, indicating a region largely depleted of carbon and containing potentially various oxidation states of silicon. Importantly, in the inset of the TEM image in figure 2(a), a lighter-contrast amorphous layer about ~5 nm thick can be distinguished just underneath the graphene. Lighter contrast possibly indicates

Figure 2. (a) Cross-sectional TEM image of the graphene/3C–SiC sample (sample 1). A ~20 nm thick amorphous layer is observed underneath the protective layer. A corresponding high magnification inset shows a carbon sheet on top of the amorphous layer. (b) and (c) The LEED patterns obtained at beam energies of 50 eV and 85 eV respectively.

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

6

lower density and/or lower carbon content, which agrees with the information from the XPS profile.

3.5. Synchrotron XPS results

Figures 5(a)–(c) show the Si 2p core level photoemission spectra of the sample measured at three different photon energies: 150 eV, 350 eV and 850 eV respectively. Note that as the photon energy is lowered, the mean free path of the photoelectrons becomes smaller, and thus the signal becomes more surface sensitive [39]. The approximate total sampling depths for individual photon energy are indicated in the figure  5. Each spectrum is deconvoluted into four Voigt doublets by a standard peak fitting routine [40–42] and the respective fitting parameters are summarized in the Supplementary Information (figure S2 and table  S1). For clarity, only the Si 2p3/2 comp onents are displayed. The binding energies, Gaussian and Lorentzian widths used for fitting are consistent with previous literature [40–42]. At 150 eV incident photon energy (figure 5(a)), a dominant peak is seen at ~101 eV, which can be attributed to Si 2p3/2 from pure SiC [43]. A clear peak resolved at ~101.6 eV, indicating the presence of either the silicon suboxide Si1+ (in the form of Si2O) [41], the oxycarbide SiOC3 [40, 44] or a mixture of both (the expected peak positions for these two environments are so close as to be not resolved here). Two comparatively weaker components corresponding to Si2+(SiO)/Si–O–C and Si4+ (SiO2) can also be seen at binding energies of ~102.3 eV and ~103.5 eV respectively. The results obtained here are in good agreement with the XPS results of Seyller et  al [40] and Virojanadara et al [45]), on the thermal oxidation of the 4H–SiC(1 1 2 0) surface. Figures 5(b) and (c) shows that these higher binding energy components decrease relative to the pure SiC signal as the incident photon energy, and therefore the total sampling depth, increases. Therefore, they can be associated with the surface region only. At 350 eV (figure

5(b)), the SiO2 component has almost disappeared, how-ever the SiOC3/Si2O and Si2+/Si–O–C components can still be observed, indicating that the silica is a superficial oxide located right at the surface. Again, this is in agreement with the results reported by Seyller et al [40]. At the highest inci-dent photon energy (850 eV), the Si2+/Si–O–C component becomes negligible, with only the SiOC3/Si2O component evident along with the SiC component. This suggests that the SiOC3/Si2O component is buried beneath the Si2+/Si–O–C layer and extends to a greater depth. It is important to note that for all three photon energies, pure SiC is the most domi-nant component, even for the most surface sensitive energy of 150 eV where the oxide components still represent a small fraction of the signal. Given the sampling depth at 150 eV is around ~2 nm, and that from the earlier ToF-SIMS and TEM measurements a ~20 nm thick amorphous oxygen-containing surface region was observed, one cannot conclude a simple model of a pure ~20 nm oxidized SiC layer on top of a pristine SiC. Instead, these results suggest a mixed phase of crystal-line SiC and amorphous oxidized SiC exhibiting a non-abrupt concentration gradient away from the surface.

Aforementioned results are consistent when we examine the carbon core level, where C1s spectra were measured at 350 eV, 850 eV and 1486.6 eV incident photon energies (main-taining approximately the same sampling depth as the silicon data) as shown in figures  5(d)–(f) respectively. At 350 eV incident photon energy (figure 5(d)) (the most surface sen-sitive), a dominant peak due to sp2-bonded carbon atoms is observed at ~284.5 eV along with that corresponding to pure SiC (~283.3 eV). The XPS curve fitting has also resolved three additional oxide-related peaks centered at 283.7 eV, 285.05 eV and 285.76 eV, indicative of the presence of SiOxCy, Si–O–C and C–O–C/Si–O–Si [40, 42, 44] respectively. The SiOxCy is an amorphous metastable phase [44] composed of Si1+, Si2+ and Si2+ components. Although the relative concentration of

Figure 3. Normalized ToF-SIMS depth profiles of selected elements from sample 1, measured to a depth of 30 nm. A significant signal from Cs2O+ ions is observed from the surface to a depth of ~20 nm indicating an oxidized SiC region toward the top of the SiC layer on Si.

Figure 4. XPS depth profile of the atomic fraction of C (black line), O (red dashed line), and Si (blue dotted line) contained in sample 1, measured to a depth of 120 nm. The first 5 nm of depth, corresponding to the graphene and top SiC layer shows a considerable amount of oxygen (35%) reducing to ~12% at 20 nm depth, and a few % beyond 120 nm depth.

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

7

each component of SiOxCy could not be determined, the prom-inent SiOC3 and Si2+/Si–O–C peaks present in the Si 2p core level spectra (figure 5(a)) indicate the predominance of the Si1+ and Si2+ components, respectively. The best fit param-eters including peak type, peak position, Lorentzian width and Gaussian width for each component are presented in the sup-plementary information (table S2). Figures 5(e) and (f) shows the intensity of the sp2 carbon peak relative to the SiC comp-onent reduces significantly when the incident photon energy is increased indicating the graphene contribution is confined to the SiC surface only as expected. The Si–O–C and Si–O–Si/C–O–C components also seem confined to the topmost part of the sample surface and start reducing relative to the SiC with an increase in the incident photon energy. However, the inten-sity of the SiOxCy component is almost constant relative to the SiC signal as the photon energy increases, confirming the oxi-dation of deeper SiC layers as observed in TEM, ToF-SIMS and XPS depth profiles. As with the Si 2p data, the pure SiC signal is quite prominent at all times; this component again supports the idea of pure SiC crystallites contained within the amorphous oxidized film.

Finally, to determine the amount of metal (Ni and Cu) resi-dues in the sample, we performed a XPS survey scans (figure S3). We noticed that the wet etch procedure was successful in removing any Ni or Cu residues completely from the sample surface.

4. Discussion (graphitization mechanism)

Considering the data presented in the preceding sections, we outline a model for the graphitization mechanism, accounting for the formation of monolayer graphene on top of a highly oxidized and amorphous region of the SiC layer. A full sche-matic describing the model is shown in figure 6.

We propose that the complete graphitization process includes the following main stages: (1) oxidation of Ni, (2) amorphization and oxidation of the top SiC layer, releasing carbon in the intermixed layer (3) graphene layer formation.

4.1. Oxidation of Ni

At low temperature (~400 °C), Ni starts reacting with the ambient oxygen present in the furnace forming Ni(II) oxide [46], XPS evidence of which was reported in our previous work [22].

4.2. Amorphization, oxidation of SiC and release of carbon

In the temperature range of 400 °C–750 °C, Ni/NiO assists in weakening the Si–C bonds and starts diffusing into the crystal-line 3C–SiC film [47]. Above ~750 °C, the Ni/NiO possibly reacts with SiC forming amorphous SiOxCy material and NiSi2 while also releasing atomic carbon [19, 48]. XPS evidence for the presence of NiSi2 before etching away the intermixed layer (NiSi2 +Ni + Cu + atomic carbon) is reported in the

Figure 5. High resolution XPS Si 2p spectra measured at three different incident photon energies: (a) 150 eV, (b) 350 eV, and (c) 850 eV. Each Si 2p spectra is deconvoluted into four Voigt doublets (SiC, Si2O/SiOC3and Si2+/Si–O–C and SiO2). Corresponding C1s core level spectra are recorded at (d) 350 eV, (e) 850 eV and (f) 1486.6 eV, with each spectrum fitted using four Voigt (SiC, SiOxCy, Si–O–C and C–O–C/Si–O–Si) and one DoniachSunjic (graphene) type peak.

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

8

supplementary information (figure S3). Some of the additional possible interfacial oxidation products formed during graphi-tization include SiO2, Si–O–C and Si–O–Si/C–O–C, evidence of which can be found in the synchrotron XPS data (figure 5).

4.3. Graphene layer formation

Note that the graphitization process takes place at 1100 °C, however, the melting point of NiSi2 and Cu is ~993 °C and ~1085 °C respectively [48], which means that the graphene growth takes place in the melt. Since the density of Cu (8.96 g cm−3) is higher than NiSi2 (4.83 g cm−3) and Ni (8.90 g cm−3) [49], at 1100 °C, the liquid Cu will likely diffuse to the bottom of the intermixed layer, as shown in figure 6(b). This molten Cu would follow conformally the surface morphology of the 3C–SiC film, acting as a pseudo Cu foil for the growth of gra-phene. The atomic carbon present in the melt is most likely catalyzed into active carbon species by Cu and subsequently chemically adsorbed potentially on both its faces upon cooling (figure 6(b)). We expect that the graphene growth herein is self-limiting, producing monolayer graphene on both top and bottom surfaces of Cu, by analogy with the more established graphene growth on copper foil [5, 50]. The intermixed layer along with top graphene and Cu layer are finally removed using the wet etch (Freckle) solution leaving bottom single layer graphene on the SiC surface (figure 6(c)). Unfortunately, this alloy-mediated process also adds undesirable roughness to the sample. The root mean square roughness of the graphene/3C–SiC sample is found around ~9 nm, which is about twofold that of bare 3C–SiC films grown on silicon substrate (~4 nm) [11].

As the graphene is obtained by means of a liquid phase epi-taxial process, the adatoms will benefit from longer diffusion lengths and thus more relaxed growth conditions as compared with the pure solid phase epitaxy when using thermal decom-position route. Overall, the nickel and the oxygen seem to play a key catalytic role in facilitating the release of C from SiC at relatively low temperatures, plus the presence of molten sili-cide enables a liquid phase epitaxy of graphene. Oxygen helps to accelerate the graphitization process as suggested in pre-vious reports [51, 52]. The presence of copper is essential for making the graphene growth more uniform, self-limiting and

controlled [23]. Note that, in addition, the catalytic action of the liquid Cu will not be hampered by grain boundaries typical in a solid Cu film/foil. Finally, we believe that the Cu is also critical for an even dilution of small amounts of nickel across large areas and for accelerating the graphitization process.

5. Conclusions

A solid-source, Ni–Cu alloy-mediated graphitization yields direct, in situ growth of a monolayer graphene on hetero- epitaxial silicon carbide on silicon substrates. Raman spectr-oscopy consistently shows an ID/IG band ratio as low as ~0.2 over large areas, indicative of the best quality graphene grown on epitaxial silicon carbide on silicon wafers reported to-date. Physical and chemical analyses suggest that the graphitiza-tion process is very likely taking place in a melt condition and includes at least three intermediate stages: (1) oxidation of Ni, (2) oxidation (and corresponding depletion of carbon) and amorphization within the top few tens of nm of the SiC layer, accompanied by the release of atomic carbon, and (3) formation of the graphene layer on the surface of the SiC layer through the key catalytic action provided by the Cu. The synergistic combi-nation of the catalytic actions of nickel and oxygen for releasing carbon from the silicon carbide, of nickel for the formation of silicides, and of copper for the self-limiting graphitization, leads to the formation of a monolayer graphene on 3C–SiC on silicon with superior uniformity and quality as compared to the use of more common thermal decomposition of 3C–SiC on Si.

The understanding of this novel approach to graphene growth and the avenues for its further improvement make it possible to finally envisage in situ, wafer-level graphene on silicon of suitable quality coating complex 2D and 3D micro/nano structures for a variety of electrical, thermal, electronic and photonic applications on silicon.

Acknowledgments

This work was performed in part at the Australian National Fabrication Facility (ANFF), at the Queensland Node and at the Victorian Node, Centre for Materials and Surface Science,

Figure 6. Schematic of the proposed alloy-mediated catalytic graphitization mechanism. (a) Double layer of Ni and Cu sputtered on the 3C–SiC surface. (b) Annealing the Cu/Ni/3C–SiC sample at 1100 °C leads to the formation of graphene layer on a ~20 nm thick oxidized SiC layer along with an intermixed layer on top. (c) The intermixed layer is removed by wet chemical etching and a graphene layer obtained directly on the 3C–SiC surface.

J. Phys. D: Appl. Phys. 50 (2017) 095302

N Mishra et al

9

La Trobe University. Synchrotron Measurements were per-formed at the Soft X-Ray Beamline at the Australian Synchro-tron. Support from the AFOSR, ONRG and Army RDECOM through the grant AOARD 15IOA053 is acknowledged. F. Iacopi is the recipient of an Australian Research Council Future Fellowship (FT120100445). The authors also acknowl-edge helpful discussions with Prof Thomas Seyller from the Technical University of Chemnitz.

References

[1] Orlita M et al 2008 Phys. Rev. Lett. 101 267601 [2] Baringhaus J et al 2014 Nature 506 349–54 [3] Lee C, Wei X, Kysar J W and Hone J 2008 Science 321 385–8 [4] Rümmeli M H, Bachmatiuk A, Scott A, Börrnert F,

Warner J H, Hoffman V, Lin J-H, Cuniberti G and Büchner B 2010 ACS Nano 4 4206–10

[5] Li X et al 2009 Science 324 1312 [6] Berger C, Song Z, Li T, Li X, Ogbazghi A Y, Feng R, Dai Z,

Marchenkov A N, Conrad E H and First P N 2004 J. Phys. Chem. B 108 19912–6

[7] Pasternak I, Wesolowski M, Jozwik I, Lukosius M, Lupina G, Dabrowski P, Baranowski J M and Strupinski W 2016 Sci. Rep. 6 21773

[8] Lee J-H, Lee E K, Joo W-J, Jang Y, Kim B-S, Lim J Y, Choi S-H, Ahn S J, Ahn J R and Park M-H 2014 Science 344 286–9

[9] Emtsev K V, Bostwick A, Horn K, Jobst J, Kellogg G L, Ley L, McChesney J L, Ohta T, Reshanov S A and Röhrl J 2009 Nat. Mater. 8 203–7

[10] De Heer W A, Berger C, Ruan M, Sprinkle M, Li X, Hu Y, Zhang B, Hankinson J and Conrad E 2011 Proc. Natl Acad. Sci. 108 16900–5

[11] Mishra N, Hold L, Iacopi A, Gupta B, Motta N and Iacopi F 2014 J. Appl. Phys. 115 203501

[12] Kermany A R, Brawley G, Mishra N, Sheridan E, Bowen W P and Iacopi F 2014 Appl. Phys. Lett. 104 081901

[13] Aristov V Y et al 2010 Nano Lett. 10 992–5[14] Ouerghi A et al 2011 Phys. Rev. B 83 205429[15] Fukidome H et al 2011 Appl. Phys. Express 4 115104[16] Gupta B, Notarianni M, Mishra N, Shafiei M, Iacopi F and

Motta N 2014 Carbon 68 563–72[17] Ouerghi A, Marangolo M, Belkhou R, Moussaoui El S,

Silly M G, Eddrief M, Largeau L, Portail M, Fain B and Sirotti F 2010 Phys. Rev. B 82 125445

[18] Suemitsu M, Miyamoto Y, Handa H and Konno A 2009 J. Surf. Sci. Nanotechnol. 7 311–3

[19] Juang Z-Y, Wu C-Y, Lo C-W, Chen W-Y, Huang C-F, Hwang J-C, Chen F-R, Leou K-C and Tsai C-H 2009 Carbon 47 2026–31

[20] Hofrichter J, Szafranek B U N, Otto M, Echtermeyer T J, Baus M, Majerus A, Geringer V, Ramsteiner M and Kurz H 2009 Nano Lett. 10 36–42

[21] Delamoreanu A, Rabot C, Vallee C and Zenasni A 2014 Carbon 66 48–56

[22] Iacopi F, Mishra N, Cunning B V, Goding D, Dimitrijev S, Brock R, Dauskardt R H, Wood B and Boeckl J 2015 J. Mater. Res. 30 609–16

[23] Cunning B V, Ahmed M, Mishra N, Kermany A R, Wood B and Iacopi F 2014 Nanotechnology 25 325301

[24] Koenig S P, Boddeti N G, Dunn M L and Bunch J S 2011 Nat. Nanotechnol. 6 543–6

[25] Iacopi F, Walker G, Wang L, Malesys L, Ma S, Cunning B V and Iacopi A 2013 Appl. Phys. Lett. 102 011908

[26] Lewis I R and Edwards H 2001 Handbook of Raman Spectroscopy: from the Research Laboratory to the Process Line (Boca Raton, FL: CRC Press)

[27] Ferrari A C and Basko D M 2013 Nat. Nanotechnol. 8 235–46[28] Ni Z H, Chen W, Fan X F, Kuo J L, Yu T, Wee A T S and

Shen Z X 2008 Phys. Rev. B 77 115416[29] Graf D, Molitor F, Ensslin K, Stampfer C, Jungen A,

Hierold C and Wirtz L 2007 Nano Lett. 7 238–42[30] Ferralis N, Maboudian R and Carraro C 2008 Phys. Rev. Lett.

101 156801[31] Stampfer C, Molitor F, Graf D, Ensslin K, Jungen A,

Hierold C and Wirtz L 2007 Appl. Phys. Lett. 91 241907[32] Yoon D, Son Y-W and Cheong H 2011 Nano Lett. 11 3227–31[33] Ferrari A C et al 2006 Phys. Rev. Lett. 97 187401[34] Fukidome H et al 2013 Proc. IEEE 101 1557–66[35] Portail M et al 2012 J. Cryst. Growth 349 27–35[36] Tiberj A, Camara N, Godignon P and Camassel J 2011

Nanoscale Res. Lett. 6 478[37] Cançado L G, Takai K, Enoki T, Endo M, Kim Y A,

Mizusaki H, Jorio A, Coelho L N, Magalhães-Paniago R and Pimenta M A 2006 Appl. Phys. Lett. 88 163106

[38] Pendry J B 1974 Low Energy Electron Diffraction: the Theory and Its Application to Determination of Surface Structure (New York: Academic)

[39] Tanuma S, Powell C J and Penn D R 1993 Surf. Interface Anal. 20 77–89

[40] Seyller T, Emtsev K V, Graupner R and Ley L 2004 Mater. Sci. Forum 457 1317–20

[41] Virojanadara C and Johansson L 2002 Surf. Sci. 505 358–66[42] Silly M, Roy J, Enriquez H, Soukiassian P, Crotti C, Fontana S

and Perfetti P 2004 J. Vac. Sci. Technol. B 22 2226–32[43] Riedl C, Coletti C and Starke U 2010 J. Phys. D: Appl. Phys.

43 374009[44] Önneby C and Pantano C 1997 J. Vac. Sci. Technol. A

15 1597–602[45] Virojanadara C and Johansson L 2004 J. Phys.: Condens.

Matter 16 S1783[46] Patnaik P 2003 Handbook of Inorganic Chemicals vol 28

(New York: McGraw-Hill)[47] Julies B A, Knoesen D, Pretorius R and Adams D 1999 Thin

Solid Films 347 201–7[48] Ward A III, Henning J P, Hagleitner H and Wieber K D

2011 Silicon-rich nickel-silicide ohmic contacts for SiC semiconductor devices US Patent No. 7,875,545 U.S. Patent and Trademark Office, Washington, DC

[49] Hammond C 2003 Handbook of Chemistry and Physics vol 2004, 84th edn (Boca Raton, FL: CRC Press) pp 39–166

[50] Zhao P, Kumamoto A, Kim S, Chen X, Hou B, Chiashi S, Einarsson E, Ikuhara Y and Maruyama S 2013 J. Phys. Chem. C 117 10755–63

[51] Bachmatiuk A, Boeckl J, Smith H, Ibrahim I, Gemming T, Oswald S, Kazmierczak W, Makarov D, Schmidt O G and Eckert J 2015 J. Phys. Chem. C 119 17965–70

[52] Mehmood F, Pachter R, Lu W and Boeckl J J 2013 J. Phys. Chem. C 117 10366–74

J. Phys. D: Appl. Phys. 50 (2017) 095302