Rapid epitaxy-free graphene synthesis on silicidated polycrystalline platinum

Abstract

Large-area synthesis of high-quality graphene by chemical vapour deposition on metallic substrates requires polishing or substrate grain enlargement followed by a lengthy growth period. Here we demonstrate a novel substrate processing method for facile synthesis of mm-sized, single-crystal graphene by coating polycrystalline platinum foils with a silicon-containing film. The film reacts with platinum on heating, resulting in the formation of a liquid platinum silicide layer that screens the platinum lattice and fills topographic defects. This reduces the dependence on the surface properties of the catalytic substrate, improving the crystallinity, uniformity and size of graphene domains. At elevated temperatures growth rates of more than an order of magnitude higher (120 μm min^-1 ) than typically reported are achieved, allowing savings in costs for consumable materials, energy and time. This generic technique paves the way for using a whole new range of eutectic substrates for the large-area synthesis of 2D materials.

ARTICLE

Received 28 Oct 2014 | Accepted 15 May 2015 | Published 15 Jul 2015

Vitaliy Babenko1, Adrian T. Murdock1, Antal A. Kos1,w, Jude Britton1, Alison Crossley1, Philip Holdway1, Jonathan Moffat2, Jian Huang3, Jack A. Alexander-Webber3, Robin J. Nicholas3 & Nicole Grobert1

Large-area synthesis of high-quality graphene by chemical vapour deposition on metallic substrates requires polishing or substrate grain enlargement followed by a lengthy growth period. Here we demonstrate a novel substrate processing method for facile synthesis of mm-sized, single-crystal graphene by coating polycrystalline platinum foils with a silicon-containing lm. The lm reacts with platinum on heating, resulting in the formation of a liquid platinum silicide layer that screens the platinum lattice and lls topographic defects. This reduces the dependence on the surface properties of the catalytic substrate, improving the crystallinity, uniformity and size of graphene domains. At elevated temperatures growth rates of more than an order of magnitude higher (120 mm min 1) than typically reported are achieved, allowing savings in costs for consumable materials, energy and time. This generic technique paves the way for using a whole new range of eutectic substrates for the large-area synthesis of 2D materials.

DOI: 10.1038/ncomms8536 OPEN

Rapid epitaxy-free graphene synthesis on silicidated polycrystalline platinum

1 Department of Materials, University of Oxford, Oxford OX1 3PH, UK. 2 Oxford Instruments Asylum Research, High Wycombe HP12 3SE, UK. 3 Department of Physics, University of Oxford, Oxford OX1 3PU, UK. w Present address: Department of Nanostructures, Research Institute for Technical Physics & Materials

Science, Budapest PO Box 49, Hungary. Correspondence and requests for materials should be addressed to N.G. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536

Graphene is a two-dimensional material that has great potential for use in a wide variety of applications such as transparent exible electrodes, photonic and electronic

devices, energy storage, sensors and coatings to name a few1. Exceptional electrical, optical and mechanical properties of the material greatly depend on the quality of the synthesized graphene that can inherently depend on the underlying substrate. Notably, the production of large area, single crystal graphene with a controlled number of layers is a challenge that has yet to be resolved.

Atmospheric pressure chemical vapour deposition (APCVD) has received much attention due to its simplicity, improved safety and lower cost over alternative procedures. When a solid substrate is used with APCVD, graphene growth is strongly inuenced by the crystallographic lattice of the substrate, its defects, roughness and grain boundaries25. Millimetre-sized single-crystal graphene akes have been synthesized recently on solid copper6 and platinum7. Time- and energy-consuming processes such as long annealing, mechanical, chemical and electrochemical polishing, re-solidication or their combinations were necessary to articially increase the substrate grain size, eliminate the grain boundaries and decrease the roughness of the surfaces. Moreover, to produce large area, single crystal, monolayer graphene with CVD, the kinetics of the synthesis must be nely balanced. Typically this is achieved with low growth rates of a few mm per minute7,8. The synthesis time is thus lengthy (hours or days) and signicant quantities of source materials are required. With unmodied commercially available polycrystalline Cu the achievable ake size is limited to only tens of micrometres9. It is also very difcult to achieve uniformity in the size and shape of graphene akes within a sample on such substrates. Consequently, graphene lms grown on solid polycrystalline metals often consist of small, irregular-sized domains with many high-angle domain boundaries, which degrade the quality of graphene10. CVD on liquid Cu gives aligned hexagonal graphene akes of irregular size distribution of up to 120 mm or regular size distribution of 2030 mm11. When these akes coalesce they often have good crystallinity and low-angle grain boundaries. However, there is little control of the nucleation and the size of such akes. The method is also difcult to upscale due to the liquidity or droplet formation.

In this investigation we combined the advantages of both approaches by using widely available, scalable, solid polycrystal-line substrates in conjunction with a thin, liquid layer on the surface of these substrates. The layer is formed by coating polycrystalline Pt foils with a silicon-containing lm, which forms a wetting liquid silicide surface in situ for graphene growth when heated, over the entire substrate surface. This signicantly reduces the inuence of the defects, roughness and the crystallographic orientations of the substrate on the quality of graphene. Substrates covered with the lm exhibit increased growth rates of up to 120 mm min 1 at elevated temperatures, one of the fastest reported to date. While metal alloys1215 and dielectric substrates16 have been shown, we believe this is the rst eutectic system used for graphene synthesis by CVD. Interestingly, studies17 show that it is possible to oxidize intercalated silicides to achieve transfer-free graphene on an insulating surface, suggesting another potential benet of the synthesis directly on metal silicides.

ResultsSubstrate preparation and large-area graphene synthesis. Pristine Pt for graphene synthesis was used as-received from the manufacturer followed by cleaning in acetone and deionized water. These substrates were not reused to avoid ongoing

accumulation of Si-impurities. To form the silicidated Pt substrates, a SiO2 lm was deposited on pristine Pt with a tetraethyl orthosilicate (TEOS) precursor in a CVD-based process, where TEOS was introduced by an Ar carrier gas (Methods). Other techniques such as evaporation and sol-gel deposition of Si/SiO/SiO2 were also investigated and showed similar trends after the optimisation of the procedure. CVD of TEOS is fast, cheap and can be performed in a single CVD system before synthesis. Upon annealing the SiO2 layer disappeared resulting in the formation of silicidated Pt (discussed later in the paper). To synthesize graphene, pristine Pt or the Pt/SiO2 stack was subjected to a CVD procedure that consisted of an annealing stage, synthesis stage and a cooling stage at atmospheric pressure (Methods). A compromise had to be made between the length of an experiment versus the size and quality of graphene akes because the ratio of H2 to CH4 needed to be kept high18 to bias the CVD reaction towards single crystal, monolayer graphene growth. In our experiments we could differentiate three graphene growth scenarios in the vast parameter space available. For example, at a temperature of 1,070 C with a hydrogen-to-methane ratio of 200:1, no growth could be observed after 30 min on silicidated Pt. A ratio of 160:1 at the same conditions resulted in the slow formation of high-quality monolayer graphene while rapid formation of multilayer graphene occurred at a H2:CH4 ratio of 100:1. The ratio of H2 to CH4 on silicidated Pt compared with pristine Pt had to be increased by as much as 50% (1,150 C) to reduce the carbon deposition and achieve akes of comparable size to those on pristine Pt. The comparison of the synthesis parameters on pristine and silicidated Pt is presented in Supplementary Fig. 1. This meant that the quality of graphene akes improved without sacricing the short experimental time. The difference in the kinetics for graphene nucleation and growth on pristine and silicidated Pt may be the key to explain this observation. Studies show that the bond strength of Si-C (B318 kJ mol 1 (ref. 19)) is higher than that of Pt-C (B225 kJ mol 1 (ref. 20)), but still lower than the graphene bond (]400 kJ mol 1 (ref. 21)) presenting a stickier silicide surface.

This suggests that the associated dwell time of carbon atoms on the silicidated surface could be prolonged, leading to the observed increase in the growth rate.

Interestingly, the high melting point of Pt allowed us to probe unexplored temperatures above 1,100 C in CVD, at which single crystal graphene akes of up to 1.8 mm were obtained in 15 min on silicidated Pt. For comparison, the window of useful H2 to

CH4 ratios on Cu is usually such that experiments often need to be conducted for hours or days8 to grow large-area, single crystal, monolayer graphene. Figure 1ac depicts the improvement in the overall size of the akes on the mm-scale that can be obtained with higher temperature on silicidated Pt. While graphene on pristine Pt showed similar trends, only smaller, polycrystalline akes of irregular shape were observed throughout the CVD parameter space (Fig. 1f). The size of graphene akes can be approximately determined by multiplying the growth rate by the synthesis time with the nucleation point separation as the maximum achievable graphene average domain size. The dependence of the growth rate as a function of temperature for silicidated and pristine Pt substrates is shown in Fig. 1d. As previously reported for Cu the nucleation density of the domains also decreased with higher temperatures22. However, on Pt the ake size was approximately an order of magnitude larger for the same temperature and time. The domain size of hexagonal or irregularly shaped graphene grown on Cu, Pt and silicidated Pt is related as follows Cu (1,050 C)oPt (1,050 C)oPt (1,150 C)osilicidated Pt (1,150 C). The difference in the nucleation point separation between pristine Pt and silicidated Pt is not as signicant as the difference in the

2 NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536 ARTICLE

1,050 C

a 1,060 C

b 1,155 C

c

d

e f

120

Growth rate, m min1

Silicidated platinum Pristine platinum

1,600

1,020 1,070 1,120 1,170

Silicidated platinum Pristine platinum

100

80

60

Nucleation point

separation

1,200

800

40

400

20

0

0

1,020 1,070 1,120 1,170

Temperature, C

Temperature,C

Figure 1 | Graphene synthesis with CVD. (ac) Representative SEM images of graphene on silicidated Pt for different temperatures: 1,050, 1,060 and 1,155 C. (d,e) Graphene growth rate and its average nucleation point separation as functions of temperature. The vertical error bars are the approximate s.d. values calculated either from different experiments or from one experiment, whichever was larger. The trend lines are guides to the eye. (f) An SEM image of monolayer graphene on pristine Pt, with many reex angles and varying edge morphology on different grains indicating polycrystallinity and strong dependence on the crystallographic orientation of Pt. Scale bars (ac) 1 mm, (f) 100 mm.

growth rate, yet it is still noticeable. The nucleation point separation (S), related to the nucleation density as p 1(S/2) 2, is also shown in Fig. 1e.

Graphene characterization and electronic properties. A number of techniques were used to extensively assess the properties of the synthesized graphene, such as size, shape, crystallinity, number of layers and its electronics properties. Figure 1c,f shows two characteristic scanning electron microscope (SEM) images of graphene akes on silicidated Pt and on pristine Pt. The latter had many reex angles that indicated numerous grain boundaries7 (Supplementary Fig. 2), and the edge morphology often varied from one Pt grain to another, similar to what has been reported by Sun et al.23 and Ping & Fuhrer24. In contrast, the hexagonal graphene akes of millimetre size grown on silicidated Pt were observed to cross sometimes hundreds of the Pt grains and their boundaries without any effect. Monolayer graphene was easily achieved on pristine Pt and silicidated Pt, most likely due to easily achieved balance regime between surface catalysis and segregation23. However, the crystallinity and size varied markedly depending on the experimental conditions. Contrary to the previous study of graphene synthesis on pristine Pt by Gao et al.7, where Pt was reused thousands of times inside a fused silica tube, we did not observe any hexagonal graphene akes on pristine Pt that was not reused. The silicide surface is conductive and not easily distinguishable under an SEM, which may explain why the presence of silicides (Supplementary Fig. 3) has not been noticed and their striking inuence on graphene growth in CVD has not been studied or reported before. The possibility to reuse Pt in combination with short experimental times can reduce costs and ensure the development of high-quality graphene syntheses commercially. We conrmed that the silicidated Pt substrates were also inert in the electrolyte solution (NaOH) that was used for a non-destructive graphene transfer (Methods) and thus could be reused.

Graphene was transferred to Si/SiO2 substrates for Raman measurements (Fig. 2a). The D-peak was almost indistinguishable from the background indicating low defect concentration. The ratio of G0 to G peaks exceeded two over large areas, as expected for good quality monolayer graphene25, both for hexagonal and irregularly shaped akes. Large-area graphene was transferred to plain transmission electron microscopy (TEM) grids (Au, Agar) without amorphous carbon support. TEM images and selected area electron diffraction patterns (SAED) were recorded. A line prole of the SAED pattern (Fig. 2b) provided additional supporting evidence for monolayer graphene26. The hexagonal shape of the akes suggested good crystallinity, but to conrm this SAED patterns were recorded at points across large areas of a graphene ake on a TEM grid (Methods, Supplementary Fig. 2). In addition, lattice-resolution AFM scans of 6 nm 6 nm were recorded on as-grown graphene

on silicidated Pt and their Fourier transforms (FT) were taken (Fig. 2c). The roughness of the substrate after solidication was limiting the resolution of AFM images, but small-area scans were sufciently at and thus suitable for the FT analysis. These methods provided ways to measure the rotation of the lattice over distances of hundreds of micrometres. The measured values showed no signicant changes in the rotation of the lattice and conrmed that the synthesized graphene was indeed a single crystal.

To assess the electronic properties of our CVD graphene a Hall-bar device (Fig. 2e) was fabricated from graphene transferred onto SiO2/Si substrates (methods) from silicidated Pt. The carrier density n and mobility m of the device were determined from the low-eld Hall measurements with n (e @rxy/@B) 1 and

m (enrxx) 1. Small p-type doping was observed in our device,

which is also commonly seen in transferred graphene on SiO2/Si substrates and considered to be due to the adsorption of water molecules between the interfaces27,28. Figure 2d shows the rxx and rxy as functions of magnetic eld from 19 to 19 T

for a typical device. We observed a very high hole mobility of 5,525 cm2 V 1 s 1 at a hole density of 5.86 1011 cm 2.

NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536

a b

c

G(2D)

G

D

Intensity, a.u.

1,100 1,500 1,900 2,300 2,700

Raman shift, cm1

d e

6

5

4

3

2

1

0

15

10

5

0

SiO2/Si

Transferred graphene

[afii9845] xx (k )

nh 5.86 1011cm2[afii9839] 5,525.4 cm2 V1s1

15

[afii9845] xy(k)

Vxx

v = 2

5

10

15

Vxy

20 15 10 5 0

B, (T)

5 10

= 2

20

Au Cr/Au

Figure 2 | Graphene characterization and electronic properties. (a) Typical Raman spectra of graphene transferred to Si/SiO2 substrates from pristine or silicidated Pt, indicating high-quality monolayer graphene. (b) SAED of the synthesized graphene with a line prole (inset) indicating monolayer graphene.

(c) Crystallinity analysis of graphene showing an SEM image of a hexagonal ake with marked areas of where lattice-resolved AFM scans were taken. An example of such a scan is shown underneath. These were performed directly on silicidated Pt. Two-dimensional Fourier transforms were taken (on the right) of the AFM scans; very small rotation of the FT patterns indicates good crystallinity over hundreds of micrometres. (d) rxx and rxy as functions of magnetic eld. The carrier density and mobility shown were determined from the low-eld Hall coefcient and the longitudinal resistance at 0 T. Also shown are the well-dened n 2 quantum Hall states with quantized rxy and vanishing rxx at |B|412 T. (e) The optical image of the actual device, where

good contrast can be observed between the transferred graphene (light blue) and the SiO2/Si substrate (dark blue), and also between the Cr/Au (yellow) and Au-only (pink) contacts. On top of the Hall bar is the ma-N 2405 resist (brown). Vxx and Vxy are also labelled. Scale bars, (c) SEM ake: 200 mm,

AFM scan: 1 nm; (e) 100 mm.

Recently, graphene has been shown to be an exceptional candidate for quantum Hall metrology29 exhibiting the highest breakdown current density of any material30. In our sample, at |B|412 T, well-dened n 2 quantum Hall states was clearly

observed with rxy h/2e2 and rxx 0, consistent with its carrier

density jBn2j nh2e. These observations conrm the high

quality of our monolayer graphene with good crystallinity, very low defects and impurities, and that graphene grown on silicidated platinum may be useful for quantum resistance metrology. The carrier mobilities obtained from the magnetotransport measurements are among the highest of reported values for CVD graphene31 on SiO2/Si substrates, and are comparable to those in epitaxial graphene grown on SiC (ref. 29). Using more sophisticated transfer methods or substrates with less charged impurities, such as h-BN32, further improvement of the carrier mobility is expected.

Substrate characterization and silicidation mechanism. It is commonly known that Pt is a noble metal that is very inert; only hot aqua regia can react with it. Silicon dioxide (SiO2), fused silica or quartz may be described similarly due to the strong SiO bond, which can be broken by hydrouoric acid or some strong bases. However, when these compounds are in contact in a reducing H2 atmosphere at elevated temperatures silica may reduce to silicon, catalysed by Pt, and react with Pt to form various phases of eutectic Pt silicides as reported by numerous studies3336. The two simplied reactions are as follows:

SiO2 2 H2 ! Si 2 H2O 1

x Si y Pt ! PtySix 2

If pure silicon is deposited onto Pt, then the reaction proceeds much quicker and starts at lower temperatures. Other lms, such as SiO, SiN also react in a similar manner. We found that a boron nitride holder was the most suitable for the Pt due to its stability in a reducing atmosphere and not-wetting by the silicide.

Characterization of the annealed and cooled substrate before graphene synthesis was performed. The cooling was rapid (quenched) to obtain the closest possible condition to that during the synthesis. Figure 3a,b shows an SEM image of a grain boundary of the substrate and its corresponding energy-dispersive X-ray (EDX) spectral map of the silicon peak. The grain boundary and the small topographic defect have higher silicon content compared with the surroundings. The silicon peak is very prominent (Fig. 3c) in such valleys, but is typically much lower, but often detectable on the Pt grains. The presence of a variety of frosted features and striations in SEM also supported the formation of a thin silicide layer (Supplementary Fig. 3). X-ray photoelectron spectroscopy (XPS) revealed that Pt silicide covered the entire substrate (see peak at around 100.5 eV (ref. 37); Fig. 3d). Additional SiO/SiO2 peaks38 were also observed and are most likely formed when the silicide was exposed to air. Complimentary X-ray diffraction (XRD) analysis was performed to conrm the chemical composition, but also to identify the phase of the solidied silicide. For this, a thick SiO2 (B3 mm) lm

was deposited on Pt to get the material in excess, resulting in the formation of a Pt3Si phase that matched well with the literature39 (Fig. 3e), but possibly other phases were present under different synthesis conditions. Cross-sectional EDX maps were recorded on Pt foils with a varying thickness of the silica layer to better understand the Pt/silica reaction and the chemical composition in

4 NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536 ARTICLE

a b c

Pt

Counts, a.u.

Si

O

0.5 1

Energy, keV

C

Pt

Pt

0

1.5 2 2.5 3 3.5 4

d e

Experimental SiO2

SiOPt silicide Overall

Intensity, a.u., log scale

Pt

Pt

Intensity, a.u.

Pt3Si or Pt2Si3

Pt3Si

Pt3Si

Pt3Si

108 105 102Binding energy (eV)

99 96

100,000

10,000

1,000

100

20

25 30 35

2 theta, degrees

40 45 50 55 60

Figure 3 | Characterization of the Pt/SiO2 substrate after annealing in hydrogen. (a) An SEM image of the annealed and cooled silicidated substrate with a lled Pt grain boundary. (b) EDX map of the Si peak, showing the valleys lled with a silicon-rich compound (green). The darker regions correspond to Pt grains with a much smaller Si-peak. (c) An EDX spectrum of the silicon-rich area from the grain boundary, where the Si peak is prominent. (d) XPS spectrum with tted peaks, showing a matching silicide peak and silicon oxide peaks. (e) XRD spectrum of the substrate (on log scale), conrming the phase as Pt3Si (monoclinic when solid). Scale bars: (a,b) 30 mm.

a b c

Figure 4 | Bulk structure of the silicidated substrates. (ac) Cross-sectional EDX maps of the silicon peak overlapped with SEM micrographs of the silicidated platinum samples. Low, medium and high silica thicknesses result in undetectable, medium and high silicide-phase presence. The silicon-containing regions and pristine Pt grains are completely indistinguishable in SEM mode. Restructuring and degradation of the foil can be observed for the samples with excess silica. Scale bars, (a) 10 mm, (b) 2 mm, (c) 50 mm.

the bulk. After the sample was annealed and quenched, regions with a thin silica layer (t500 nm) did not show a detectable silicon peak in the EDX maps (Fig. 4a). Regions with a higher initial silica thickness (4500 nm) showed a high silicon signal (Fig. 4b) around the grain boundaries, suggesting that the reaction proceeded faster in those areas while progressing slowly into the bulk of the grains. Finally, excess silica (\3,000 nm)

resulted in the restructuring and degradation of the foil due to the formation of a two-phase composite (Fig. 4c), held together by the silicon-containing compound.

On the basis of experimental observations of the SiO2 lm before and after annealing we propose a simple mechanism of the liquid silicide formation that explains why the morphologies of graphene are so different on pristine and silicidated Pt. Most Pt silicide phases have reduced (eutectic) melting points40, especially in the form of a thin lm. For example, the identied monoclinic Pt3Si phase melts at around 830 C (ref. 40). Figure 5 shows a diagram of our proposed silicide formation mechanism and the corresponding SEM images of the Pt/SiO2 stack. The deposited

silica layer is not conductive and appears dark under the electron beam. Upon heating, a spontaneous reaction occurs at a variety of points in the lm that start to spread in a circular manner. We believe the circular holes in the silica lm are lled with a thin layer of a conductive liquid, which when cooled appear similar to the pristine Pt but with a frosted surface (Fig. 5b,c, Supplementary Fig. 3). The silicide evaporates from the surface, diffuses into the bulk and locally smoothes any topographical defects, such as Pt grain boundaries, as inferred from the cross-sectional EDX maps in Fig. 4. It should be noted that the annealing time becomes an important parameter. Graphene synthesis can be started shortly after the SiO2 thin lm completely reacts with Pt and the surface has a uniform chemical composition. We do not rule out chemical inuences to the reaction mechanism by the silicidated Pt surface in addition to simple mechanical (liquid) effects. For example, the silicide route may produce silicon carbide as an intermediate state. Precise in situ studies would be needed to conrm the correct chemical mechanism. We also suggest that such eutectic materials

NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536

may be suitable for use as liquid-surface substrates at low synthesis temperatures, bringing all the associated benets.

Graphene crystallinity control. Good crystallinity is an important quality factor of CVD-grown graphene, as it offers superior electronic and tensile properties, chemical resistivity and impermeability10. The ability to control the crystallinity allows us to improve the material. A systematic investigation into the effect of the thickness of the deposited silica layer on the hexagonality of graphene akes was conducted. We used a simple analysis technique to distinguish between different levels of hexagonality of the synthesized samples. Image processing was used to obtain a threshold to convert SEM micrographs into binary images of

graphene domains (Fig. 6ad). Each solid shape was then characterized individually with a computer script such that its pixel area and perimeter were measured to obtain a value relative to a perfect hexagon as detailed in the Methods section. The results are presented in Fig. 6e. The trend matches well with the expectation: the more silica is added, the thicker the liquid silicide layer is and thus the akes grow more independently from the underlying solid substrate. The value of the hexagonality saturates at 800-nm-thick silica layer with a 91% similarity to a perfect hexagon. Another important quality indicator, the uniformity of the shapes, also decreases signicantly from about 50% uctuation at 0 nm to less than 5% at 800 nm, as shown by the error bars. Usable silica thicknesses are limited by the thickness of the Pt foil. If there is too much silica (42 mm) the 25 mm Pt foil can start to degrade (Fig. 4c) or curl during the synthesis due to thermal stress caused because of the differences in the thermal expansion coefcients of Pt and silica. In addition, the silicide layer may undergo a phase change during cooling resulting in the formation of striations that may induce aligned cracks in graphene (Supplementary Fig. 3). A thin silica lm is thus benecial to avoid the formation of tall striations.

Although there is as yet no conrmed theory that explains the poor crystallinity of graphene on pristine Pt via CVD that we have observed, it is well known that the face-centred cubic (FCC) lattice of Pt has the largest lattice constant compared with the commonly used transition metals for graphene synthesis. On Cu(111) the lattice constant mismatch with graphene is 3.6%, whereas on Pt(111) this is signicantly higher at 12.5%. This may result in incommensurable growth and the observed poor crystallinity.

DiscussionIn this study, we demonstrate a novel, low-cost substrate processing procedure to achieve rapid, efcient synthesis of millimetre-sized single crystal graphene. Silicidation of poly-crystalline foils may be used as a replacement for expensive single-crystal substrates or as an alternative to polishing and grain enlargement. Large topographic defects, such as grain boundaries, are locally smoothed and the lattice of the grains is screened by a liquid surface, making silicidated polycrystalline foils superior to single crystals due to epitaxy-free synthesis. This ensures that graphene grows as a single crystal in its undistorted hexagonal shape. Silicidated polycrystalline Pt also exhibits low graphene ake nucleation density of up to 0.3 mm 2 and enhanced growth rates of up to 120 mm min 1 at elevated temperatures, allowing

a

SiO2

SiO2

Pt

Pt

Pt3Si

b

Pt3Si

c

Pt

Figure 5 | Stages of the silicide formation. (a) The substrate is coated with a SiO2 lm. (b) Upon annealing a spontaneous reaction occurs resulting in circular holes in the lm. (c) A thin liquid silicide layer covers the substrate and lls in any valleys or Pt grain boundaries. All the scale bars in the complimentary SEM micrographs are 1 mm.

a b e

100

75

50

25

0

0

Hexagonality, %

c d

400 800 1,200 1,600 2,000

Silica film thickness, nm

Figure 6 | Inuence of the silica layer thickness on the hexagonality and thus the crystallinity of the CVD graphene. (a,b) A sample with a high hexagonality and uniformity of graphene domains on silicidated Pt. A SEM micrograph is shown with a processed binary image of the resulting shape distribution. (c,d) A sample with low hexagonality and shape uniformity on pristine Pt. (e) With increased silica layer thickness the crystallinity and the shape uniformity increases. Error bars represent the distribution of shapes on the sample (Methods). The dashed line is a guide to the eye. Scale bars, (a) 400 mm, (c) 200 mm.

6 NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536 ARTICLE

millimetre-sized akes to grow easily in minutes. The silicide lm formation on the substrates is an important parameter for graphene growth and may have benecial effects if controlled. To our knowledge no current reports focusing on graphene synthesis via CVD consider the effects of the formation of the silicides, though commonly used reaction tubes are made of fused silica. It is likely that previous results were inuenced by these silicides but were overlooked by other reports. In particular, we explicitly verify the dependence of the amount of silica on the crystallinity of graphene akes on Pt. All of these benecial effects may make the silicidation feasible commercially, especially if the procedure is extended to other metals. There is signicant potential for further use of the method, such as a liquid surface for low temperature graphene synthesis or the possibility to oxidize the silicide layer to achieve transfer-free graphene. This approach is generally applicable and paves the way for a whole new eld of eutectic substrates for the growth of superior quality 2D materials.

Methods

Substrate preparation. Pristine Pt substrates (25 mm, Goodfellow, 99.95% purity) were used as received from the supplier after cleaning in acetone and deionized water. To form a Pt/SiO2 stack, a lm of SiO2 of specied thickness (Fig. 6e)

was deposited onto pristine Pt via CVD with a tetraethyl orthosilicate (TEOS) precursor. For example, a SiO2 lm of about 1 mm thickness can be depositedon pristine Pt with CVD by passing Ar gas (350 sccm) through a TEOS precursor (in a bubbler) that is carried to a furnace with the substrates heated to 700 C for 30 min. Other methods, such as Si, SiO, SiO2 evaporation, sputter coating, sol-gel processes or ion implantation are also applicable.

Graphene synthesis. Pristine Pt or the Pt/SiO2 stack was placed on a boron nitride holder inside a fused silica tube (28 mm inner diameter) at sufcient distance from the walls of the tube. The tube was then purged with Ar (99.999% purity) and shifted into a pre-heated furnace, followed by annealing in H2(99.995%, 200 sccm) to allow Pt silicide formation and to stabilize the temperature of the furnace. The useful annealing time to form silicidated Pt strongly depended on the temperature and the silica thickness. For example, at 1,070 C with 1-mm-thick silica layer, 30 min was suitable (typically used, also for higher temperatures). However, at 1,050 C with a 1 mm silica lm after 30 min, 10% of the SiO2 lm still remained. In the experiments where reusing silicidated Pt was tested, we shortened the annealing time to 5 min to simply re-heat the solidied silicide. To synthesize graphene, a mixture of CH4 (99.5%) and H2 was introduced. The gas ratio and the duration of the synthesis depended greatly on the temperature and the silica layer thickness and are given in the Supplementary Fig. S1 for both pristine and silicidated Pt. For example, for a 1-mm-thick silica layer at 1,070 C the ow rates were set to 4 sccm CH4 and 600 sccm H2 for 20 min, resulting in B0.5-mm-sized graphene akes. Directly after this step, the tube was shifted from the furnace and purged with Ar to quench graphene growth.

Transfer to Si/SiO2 substrates and TEM grids. The bubbling transfer method outlined in Gao et al.7 was employed. The sample was spin-coated (500 r.p.m.) with polymethylmethacrylate (4% by weight PMMA in chlorobenzene, 996,000 Mw, Sigma-Aldrich) and submerged into an electrolyte solution (1 M NaOH). A potential difference was applied to the electrodes (6 V, Pt/graphene/PMMA on the negative terminal) submerged into the solution. This resulted in bubbles forming between the PMMA and Pt, separating them. PMMA with a graphene layer was then rinsed in deionized water and lifted onto a Si/SiO2 wafer, followed by heating at 140 C for 5 min to improve the adhesion of graphene to SiO2. The stack was then placed in acetone (99.99%, Sigma-Aldrich) for 6 h and after dried with a N2 gun. For the TEM transfers, the PMMA layer was burnt in air at 350 C after the transfer to a gold TEM grid, as outlined in Huang et al.41

Graphene characterization. Raman spectra and maps were recorded with a Horiba LabRAM Aramis microscope equipped with a 532 nm laser. Routine SEM characterization was done with a JEOL JSM-6500F microscope at 5 kV. TEM images and SAED patterns were recorded with a JEOL JEM-2010 microscope at 80 kV. AFM measurements were performed using a Cypher AFM (Asylum Research). The scan was done in contact mode by looking at the lateral motion due to changes in friction as the probe scanned over the surface. To eliminate the effects of the direction of probe movement and also any topographic effects the trace scan was subtracted from retrace for the lattice-resolved image in real space. 2D fast Fourier transforms were taken of the AFM scans, the absolute values of which were then calculated and plotted on log scale; the contrast of the nal images was adjusted to reduce the background.

Electronic properties. Eight-leg Hall bars (15 mm 130 mm) were fabricated

using e-beam lithography followed by oxygen plasma etching. A two-step ohmic contacting method42 was used with thermally evaporated Cr/Au outer contacts and Au-only inner contacts (Fig. 2e). The nal devices were coated in ma-N 2405 photoresist as a dielectric layer and a corona discharge technique43 was employed to control the charge carrier density. Electrical measurements were carried out at1.4 K in an Oxford Instruments 21 T superconducting magnet, where magnetic elds B were applied perpendicular to the sample surface. A small AC current with peak amplitude of 40 nA was applied in the longitudinal direction (Ixx). The

longitudinal voltage Vxx and the Hall voltage Vxy (Fig. 2e) were simultaneously

measured using the standard lock-in technique. The longitudinal resistivity rxx and

the Hall resistivity rxy were determined as (W/L)?Vxx/Ixx and Vxy/Ixx, respectively, where L and W are the length and the width of the device.

Substrate characterization. The thickness of the SiO2 lms was measured with a Nanometrics Nanospec AFT interferometer on co-deposited Si tiles. Some samples were also analysed with ellipsometry directly on a Beaglehole Instruments Picometer Ellipsometer. The XPS measurements were performed at base pressure of 5 10 10 torr on a VG nine channel CLAM4 electron energy analyser with a

250 W Al X-ray excitation. XRD analysis was performed on the Bruker D5000 diffractometer.

Hexagonality analysis. MATLAB was used to nd an automatic threshold value of SEM micrographs, which were then converted into binary images. The ratio of the perimeter (P) squared to area (A) was calculated for each geometric shape. This unit-less number is usually indicative of specic geometric shapes. For example, for a circle it is 4pE12.57 (the lowest possible value), for an octagon it is E13, for a hexagon O192E13.86, for a pentagon E15, for a square it is 16 and so on.

Concave polygons have higher P2/A values. If the calculated ratio (P2/A) for a shape is X, then we dene the hexagonality as E1 |X O192|/O192 (or as a

percentage, multiplied by 100%). This means that a shape with a value close to 100% is close to a perfect hexagon. The s.d. of the % hexagonality for a sample indicates how non-uniform the shape distribution is. It is understood that rounded convex or concave polygons may introduce an uncertainty in the measurement, since, for example, a rounded cube may have the ratio of the perimeter squared to area similar to a hexagon. However, we usually have either hexagonal shapes (silicidated Pt) or concave polygons (pristine Pt); thus, for a large number of graphene domains the approach characterizes the samples sufciently.

References

1. Novoselov, K. S. et al. A roadmap for graphene. Nature 490, 192200 (2012).2. Wood, J. D., Schmucker, S. W., Lyons, A. S., Pop, E. & Lyding, J. W. Effects of polycrystalline Cu substrate on graphene growth by chemical vapor deposition. Nano Lett. 11, 45474554 (2011).

3. Zhao, L. et al. Inuence of copper crystal surface on the CVD growth of large area monolayer graphene. Solid State Commun. 151, 509513 (2011).4. Ogawa, Y. et al. Domain structure and boundary in single-layer graphene grown on Cu(111) and Cu(100) lms. J. Phys. Chem. Lett. 3, 219226 (2011).

5. Murdock, A. T. et al. Controlling the orientation, edge geometry, and thickness of chemical vapor deposition graphene. Acs Nano 7, 13511359 (2013).

6. Wu, T. et al. Triggering the continuous growth of graphene toward millimeter-sized grains. Adv. Funct. Mater. 23, 198203 (2013).

7. Gao, L. B. et al. Repeated growth and bubbling transfer of graphene with millimetre-size single-crystal grains using platinum. Nat. Commun. 3, 699 (2012).

8. Zhou, H. et al. Chemical vapour deposition growth of large single crystals of monolayer and bilayer graphene. Nat. Commun. 4, 2096 (2013).

9. Yu, Q. et al. Control and characterization of individual grains and grain boundaries in graphene grown by chemical vapour deposition. Nat. Mater. 10, 443449 (2011).

10. Tapaszt, L. et al. Mapping the electronic properties of individual graphene grain boundaries. Appl. Phys. Lett. 100, 053114 (2012).

11. Geng, D. C. et al. Uniform hexagonal graphene akes and lms grown on liquid copper surface. Proc. Natl Acad. Sci. USA 109, 79927996 (2012).

12. Weatherup, R. S. et al. In situ characterization of alloy catalysts for low-temperature graphene growth. Nano Lett. 11, 41544160 (2011).

13. Dai, B. Y. et al. Rational design of a binary metal alloy for chemical vapour deposition growth of uniform single-layer graphene. Nat. Commun. 2, 522 (2011).

14. Shin, H. A. S. et al. Highly uniform growth of monolayer graphene by chemical vapor deposition on Cu-Ag alloy catalysts. Phys. Chem. Chem. Phys. 16, 30873094 (2014).

15. Yan, K., Fu, L., Peng, H. L. & Liu, Z. F. Designed CVD growth of graphene via process engineering. Acc. Chem. Res. 46, 22632274 (2013).

16. Sun, J. Y. et al. Direct growth of high-quality graphene on high-kappa dielectric SrTiO3 substrates. J. Am. Chem. Soc. 136, 65746577 (2014).

17. Lizzit, S. et al. Transfer-free electrical insulation of epitaxial graphene from its metal substrate. Nano Lett. 12, 45034507 (2012).

18. Vlassiouk, I. et al. Role of hydrogen in chemical vapor deposition growth of large single-crystal graphene. Acs Nano 5, 60696076 (2011).

NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8536

19. Guselnikov, L. E. & Nametkin, N. S. 1,1-Dimethyl-1-silaethylene - heat of formation, ionization-potential and the energy of the silicon-carbon pi-bond.J. Organomet. Chem. 169, 155164 (1979).20. Kua, J., Faglioni, F. & Goddard, W. A. Thermochemistry for hydrocarbon intermediates chemisorbed on metal surfaces: CHn-m(CH3)(m) with n 1,

2, 3 and mo n on Pt, Ir, Os, Pd, Ph, and Ru. J. Am. Chem. Soc. 122,

23092321 (2000).21. Long, L. H. & Norrish, R. G. W. The thermochemistry of carbon - valence states, heats of sublimation and energies of linkage. Proc. R. Soc. London Ser-A 187, 337357 (1946).

22. Vlassiouk, I. et al. Graphene nucleation density on copper: fundamental role of background pressure. J. Phys. Chem. C 117, 1891918926 (2013).

23. Sun, J. et al. Growth mechanism of graphene on platinum: Surface catalysis and carbon segregation. Appl. Phys. Lett. 104 (2014).

24. Ping, J. L. & Fuhrer, M. S. Carbon impurities on graphene synthesized by chemical vapor deposition on platinum. J. Appl. Phys. 116 (2014).

25. Malard, L. M., Pimenta, M. A., Dresselhaus, G. & Dresselhaus, M. S. Raman spectroscopy in graphene. Phys. Rep. 473, 5187 (2009).

26. Meyer, J. C. et al. On the roughness of single- and bi-layer graphene membranes. Solid State Commun. 143, 101109 (2007).

27. Moser, J., Verdaguer, A., Jimenez, D., Barreiro, A. & Bachtold, A. The environment of graphene probed by electrostatic force microscopy. Appl. Phys. Lett. 92, 123507 (2008).

28. Lafkioti, M. et al. Graphene on a hydrophobic substrate: doping reduction and hysteresis suppression under ambient conditions. Nano Lett. 10, 11491153 (2010).

29. Tzalenchuk, A. et al. Towards a quantum resistance standard based on epitaxial graphene. Nat. Nanotechnol. 5, 186189 (2010).

30. Alexander-Webber, J. A. et al. Phase space for the breakdown of the Quantum Hall effect in epitaxial graphene. Phys. Rev. Lett. 111, 096601 (2013).

31. Mattevi, C., Kim, H. & Chhowalla, M. A review of chemical vapour deposition of graphene on copper. J. Mater. Chem. 21, 33243334 (2011).

32. Gannett, W. et al. Boron nitride substrates for high mobility chemical vapor deposited graphene. Appl. Phys. Lett. 98, 242105 (2011).

33. Lamber, R. & Jaeger, N. I. On the reaction of Pt with SiO2 substrates: Observation of the Pt3Si phase with the Cu3Au superstructure. J. Appl. Phys. 70, 457461 (1991).

34. Wang, D. et al. Silicide formation on a Pt/SiO2 model catalyst studied by TEM, EELS, and EDXS. J. Catalysis 219, 434441 (2003).

35. Penner, S. et al. Platinum nanocrystals supported by silica, alumina and ceria: metalsupport interaction due to high-temperature reduction in hydrogen. Surf. Sci. 532535, 276280 (2003).

36. Tsui, B.-Y. & Chen, M.-C. Dielectric degradation of Pt/SiO2/Si structures during thermal annealing. Solid-State Electronics 36, 583593 (1993).

37. Grunthaner, P. J., Grunthaner, F. J. & Madhukar, A. Chemical bonding and charge redistribution - valence band and core level correlations for the Ni/Si, Pd/Si, and Pt/Si systems. J. Vac. Sci. Technol. 20, 680683 (1982).

38. Nguyen, T. P. & Lefrant, S. XPS study of SiO thin-lms and SiO metal interfaces. J. Phys. Condens. Mater. 1, 51975204 (1989).

39. Chou, T. C. Anomalous solid-state reaction between SiC and Pt. J. Mater. Res. 5, 601608 (1990).

40. Xu, L. L., Wang, J., Liu, H. S. & Jin, Z. P. Thermodynamic assessment of the PtSi binary system. Calphad 32, 101105 (2008).

41. Huang, P. Y. et al. Grains and grain boundaries in single-layer graphene atomic patchwork quilts. Nature 469, 389392 (2011).

42. Baker, A. M. R., Alexander-Webber, J. A., Altebaeumer, T. & Nicholas, R. J. Energy relaxation for hot Dirac fermions in graphene and breakdown of the quantum Hall effect. Phys. Rev. B 85, 115403 (2012).

43. Lartsev, A. et al. Tuning carrier density across Dirac point in epitaxial graphene on SiC by corona discharge. Appl. Phys. Lett. 105, 063106 (2014).

Acknowledgements

The work at Oxford University Materials Department is supported by the DTA-EPSRC(V.B., J.B.), the Commonwealth Scholarship Commission (A.T.M.), University of Oxford Clarendon Fund (A.T.M.), the Royal Society (N.G.), the European Research Council (ERC-2009-StG-240500) (N.G.), the Engineering and Physical Sciences Research Council (N.G.). The work at Oxford University Physics Department is supported by the EPSRC.

Author contributions

N.G., V.B. conceived and designed the investigation. V.B. carried out the CVD synthesis and optimization, transfer experiments, TEM characterization, data analysis and wrote the manuscript. V.B., A.T.M., A.K., J.B., performed SEM characterization. V.B., A.K. performed Raman characterization. N.G. oversaw project direction, planning and infrastructure. A.C. carried out the XPS measurements. P.H. carried out the XRD measurements. J.M. performed the AFM measurements. J.A.A.-W., J.H. and R.J.N. carried out the Quantum Hall device fabrication, measurements and data analysis. All authors discussed the results and commented on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Babenko, V. et al. Rapid epitaxy-free graphene synthesis on silicidated polycrystalline platinum. Nat. Commun. 6:7536 doi: 10.1038/ncomms8536 (2015).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

8 NATURE COMMUNICATIONS | 6:7536 | DOI: 10.1038/ncomms8536 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.