Growth and Intercalation of Graphene on Silicon Carbide Studied by Low-Energy Electron Microscopy
Abstract
Based on its electronic, structural, chemical, and mechanical properties, many potential applications have been proposed for graphene. In order to realize these visions, graphene has to be synthesized, grown, or exfoliated with properties that are determined by the targeted application. Growth of so-called epitaxial graphene on silicon carbide by sublimation of silicon in an argon atmosphere is one particular method that could potentially lead to electronic applications. In this contribution we summarize our recent work on different aspects of epitaxial graphene growth and interface manipulation by intercalation, which was performed by a combination of low-energy electron microscopy, low-energy electron diffraction, atomic force microscopy and photoelectron spectroscopy.
1 Introduction
Since the groundbreaking work by Novoselov and Geim and coworkers,1, 2 graphene has raised the interest of physicists and chemists as well as engineers from various disciplines. The interest is caused by graphene's unique physical properties which suggest that it could lead to novel and improved applications which cover different aspects such as electronic devices, energy storage, composite materials and many more.3, 4 One of the primary objectives of graphene research is its synthesis with special emphasis on high quality, large area, and high volume.
Among the different techniques studied in the past, the epitaxial growth of graphene on silicon carbide (SiC) substrates appears to be a highly promising method for the development of electronic devices like, e.g., high frequency transistors,5-7 frequency mixers,8 THz detectors,9 and many more. In fact, the use of so-called epitaxial graphene10 on SiC has already been suggested by Berger and de Heer in the early days of graphene research.11, 12 The process for growing epitaxial graphene on SiC is conceptually very simple. At temperatures of 1150°C and above, SiC starts to decompose at the surface. While Si atoms sublimate from the surface, C atoms can form stable bonds thus staying behind at the surface. The stable modification of carbon under the given conditions is graphite. Hence, graphene or graphite is formed at the surface, depending mainly on the annealing temperature. The formation of thin graphitic films upon annealing of SiC in vacuum has already been observed early on in surface studies of SiC by van Bommel et al.13 and was later characterized in more detail by Forbeaux and coworkers.14, 15 For this reason, the growth technique is generally referred to as sublimation growth. An advantage of this technique is that SiC wafers are nowadays available with a diameter of up to 6 inches. Intrinsic or compensated SiC, so-called semi-insulating SiC, has such a low carrier concentration that it is basically an insulator. This offers the advantage to grow graphene on a wafer scale directly on an insulating substrate.
Sublimation growth of graphene on SiC depends strongly on the conditions and the orientation of the SiC surface.10 On the Si-face, the (0001) surface of hexagonal polytypes, sublimation growth can be controlled much easier than on the C-terminated 
surface. In addition, on SiC(0001) a fixed orientation exists for the grown graphene layers whereas this is not the case for SiC
. Therefore, many groups prefer to grow graphene on the Si-face. If sublimation growth is carried out in vacuum (typically ultrahigh vacuum, UHV), the graphene films are comprised of small grains on a rough surface. A breakthrough was achieved when the growth was carried out in an Ar atmosphere16, 17 which slows down Si sublimation from the surface. This technique leads to considerably flatter surfaces with graphene domains limited by the step structure of the SiC(0001) surface caused by an unavoidable slight misorientation of the wafer (see also below) and it is widely used in the community today for growing epitaxial graphene on SiC(0001). Alternatively, the so-called confinement controlled sublimation can be employed, where the substrate is annealed in an enclosure.18 Using graphene produced on SiC(0001) by sublimation growth in Ar allowed the observation of the chiral quantum Hall effect19, 20 which indicates negligible perturbations of the electronic structure of epitaxial graphene.
The progress in the growth of graphene enabled research in new directions concerning fundamental aspects of the material. An important aspect is the role of the interface. For example, the buffer layer with 
periodicity that exists at the interface between epitaxial graphene and the SiC(0001) surface, gives rise to the observed n-type doping of epitaxial graphene.21, 22 Since the buffer layer itself is a covalently bound carbon layer with an atomic arrangement identical to graphene,23 it seems likely that it can be lifted off from the substrate by intercalation as was previously done with graphene (called monolayer graphite) on Ni(111).24 This has led to many interesting experimental investigations, in which the intercalation of many different elements like, e.g., hydrogen,25 lithium,26 oxygen,27 fluorine,28 germanium,29 copper,30 or gold31 has been studied. On the other hand, because the buffer layer is only observed on SiC(0001)23 and not on SiC
, one might ask the question if graphene can be grown on other surface orientations of SiC such as, for example, the 
or 
surfaces of hexagonal SiC polytypes and what the properties of the layers would be.
Low-energy electron microscopy (LEEM) is an ideal tool to study growth and intercalation of graphene.32 In the case of epitaxial graphene on SiC, this was initially demonstrated by Hibino et al.33 and Ohta et al.,34 who studied graphene formation and by Riedl et al.25 who investigated hydrogen intercalation. Hence we have employed this powerful tool to investigate several different aspects of epitaxial graphene growth and intercalation. The present paper gives an overview of these results, which partially have been published before.35, 36 It is organized as follows. General experimental aspects of the studies are described in section 2. Section 3 discusses the influence of adding silane to the Ar atmosphere during sublimation growth on the morphology of the surface, followed by a LEEM dark field characterization of the graphene layer in section 4. In section 5 we investigate the growth of graphene on 4H-SiC
and 4H-SiC
.35 While section 6 is concerned with the process of decoupling the buffer layer by interface oxidation through annealing in water vapor,36 section 7 introduces first results concerning the initial stages of hydrogen intercalation under the buffer layer. Finally, section 8 gives a brief summary.
2 General Experimental Aspects
The preparation of epitaxial graphene was performed using previously described techniques.37 For short, a custom-built, computer-controlled reactor was used, that allows annealing of SiC samples in Ar environment (typically at a pressure of 
bar) to temperatures of up to 2000°C by inductive heating of a graphite susceptor. In this way, SiC(0001) substrates covered by either the buffer layer (
) only or covered with a monolayer of epitaxial graphene (MLG) can be obtained routinely. Note, that MLG resides on top of the buffer layer. The respective annealing temperatures were 
C and 1675°C. The annealing time was 15 min. Prior to that, the samples were etched in molecular hydrogen to remove polishing scratches. Unless otherwise stated, nominally on-axis, n-type as well as semi-insulating 6H-SiC(0001) substrates from SiCrystal and II-VI-Inc., respectively, were used.
The samples were characterized by X-ray photoelectron spectroscopy (XPS) in the home laboratory using monochromated Al Kα radiation provided by a Specs XR 50 X-ray source in conjunction with a Specs Focus 500 monochromator and a Specs Phoibos 150-MCD analyzer. Angle-resolved photoelectron spectroscopy (ARPES), which is not explicitly discussed in the present paper, was performed at BESSY II using either a Specs Phoibos 100-2D-CCD analyzer or a toroidal electron spectrometer.38 Low-energy electron microscopy (LEEM), photoemission electron microscopy (PEEM) and low-energy electron diffraction (LEED) was performed in the home laboratory with a Specs FE-LEEM P90 system. By acquiring a series of LEEM images with the electron energy varied in equidistant steps, one can extract the reflected electron intensity as a function of the energy for each pixel of the series. Thus, low-energy electron reflectivity (LEER) spectra are obtained. Atomic force microscopy (AFM) images were obtained with the help of a Park Scientific XE100 microscope. Micro-Raman measurements of water-treated samples were obtained with a Jobin Yvon T64000 triple spectrometer under ambient conditions. A 100× objective was used to focus a frequency-doubled Nd:YVO4 laser (wavelength 532 nm) on the sample.
3 Influence of Silane on Atmospheric Pressure Graphitization
The advantage of sublimation growth of graphene on SiC in Ar16, 17 is that it slows down Si sublimation so that higher growth temperatures can be employed which improves the surface morphology. In a similar fashion, the annealing of SiC in a partial pressure of disilane can be employed.39-41 These reports have motivated us to investigate in how far the addition of silane to the Ar atmosphere used during sublimation growth leads to modifications of the surface morphology as compared to the standard Ar process.
To that end, silane was added to the pure Ar flow used in the standard process in the form of a mixture of 0.1% SiH4 in Ar via a separate flow controller. The two processes described here, denoted ‘silane 1’ and ‘silane 2’, differ mainly in the flow rate of the SiH4/Ar mixture, which is five times larger for the silane 2 process. The temperature of 1800°C has been optimized to yield an average graphene thickness of one monolayer as determined by XPS for the process silane 1 for small off-axis angles (see below). The temperature was held for 15 minutes. The process parameters are summarized in Table 1. The higher temperature needed indicates that the Si sublimation from the surface is further suppressed by the silicon available in the gas phase.
| Process | Ar flow (slm) | SiH4 flow (slm) | Temperature (°C) |
|---|---|---|---|
| Argon | 0.1 | – | 1675 |
| Silane 1 | 0.1 | 0.01 | 1800 |
| Silane 2 | 0.15 | 0.05 | 1800 |
The topography of the samples after the different processes was investigated by AFM, and typical results are shown in Figure 1(a)–(d). The terrace structure visible in Figure 1(a), although somewhat less regular than in Ref. [16], is typical of a 6H-SiC(0001) sample after graphitization in argon at 1675°C in our hot-wall reactor setup. Images after graphene growth using the silane 1 and silane 2 process are displayed in Figures 1(b) and (c), respectively. In both images, well-developed terraces can be seen which have a considerably larger width as compared to the standard sample (note that the images are on the same scale). From the AFM line profiles in Figure 1(d), which have been extracted along the lines indicated in the micrographs, a clear increase in the step heights for the silane processes is obvious. This suggests that the wider terraces are the result of a more pronounced step bunching than in the case of annealing in pure argon.16 Note that both step heights and terrace widths are larger for the process with the higher silane flow rate.
However, the terrace width is not only influenced by the conditions of graphitization, but it also depends on the actual surface orientation of the nominally on-axis oriented 6H-SiC(0001) wafers.42, 43 Therefore, we studied the morphologies obtained by the different processes discussed above systematically in dependence of the deviation of the surface normal from the (0001) direction, which is expressed in terms of the miscut angle φ. The miscut varied between 
and 0.85° on the samples used in this study.
For hydrogen etched surfaces, which are the starting point of our graphitization processes, the relation between terrace width and miscut is simple. Typically step heights of one or one half of a 6H unit cell, corresponding to 1.5 nm and 0.75 nm respectively, are observed for the process employed here.37 For geometric reasons, the terrace width w after hydrogen etching is then given by 
when half unit cell steps are assumed. After graphitization however, the degree of step bunching varies, so that AFM was used to estimate local miscut angles and the average terrace widths after graphitization.
The results are compiled in Figure 1(e) for the standard argon process (green downward triangles), the silane 1 (blue dots) and the silane 2 (red upward triangles) processes. For comparison, the trend for hydrogen etched surfaces as described above is also shown. In general, the terrace width decreases with increasing miscut angle. This is in agreement with earlier studies where an inverse proportionality was found for graphene growth on SiC(0001) in argon, albeit at a much lower pressure in the mTorr range .43 Closer inspection shows that after the graphitization the observed terrace widths w can be empirically described as a power law of φ. The results of fits to the data, included as solid lines in Figure 1(e), serve as guides to the eye. For all miscut angles investigated, the terrace width obtained by addition of silane exceeds the width obtained with the standard process. For the silane 1 process, the step bunching as judged from step heights observed in AFM is stronger by a factor of 1.4…2.3, with the largest increase at higher angles. For the silane 2 process, an increase of 2.3…6.0 is found that is most pronounced for small miscuts. Best results for the latter were observed at around 
with mostly straight terraces, whereas a more irregular morphology was observed for even smaller miscuts.
Homogeneity and reproducibility of the graphene coverage are additional factors of prime importance to obtain large area graphene on SiC(0001). In a first step, the graphene thickness was determined by XPS measurements as described previously37, 44 with the analysis spot centered on the sample. The local miscut angle was averaged over several AFM images acquired in the XPS analysis region. For some samples, scanning electron microscopy (SEM) as a more local way to estimate the graphene thickness at the positions of the AFM images was used, similar to what has been reported in the literature.45 The graphene coverage thus determined for samples subjected to the different processes is plotted as a function of the miscut angle in Figure 1(f). First it should be noted that there is a reasonable agreement between the values derived from the local SEM data for the silane 1 process and from XPS measurements which average over a larger area. Second, the same dependence of the graphene thickness on the angle is observed for the silane 1 and the standard process. Empirical fits by a power law are included in Figure 1(f) as guides to the eye. Towards smaller miscuts, where the terrace width rapidly increases [see Figure 1(e)], there is a steep decrease of the coverage. Towards higher miscuts, the terrace width slowly decreases and the coverage slowly increases. For both terrace width and graphene thickness, the transition between these two regimes occurs at about 
. This points towards an influence of the terrace width on the graphene thickness. Indeed, it was found that graphene growth nucleates at step edges of the SiC surface, where typically the formation of bi- and trilayer graphene is seen on samples dominated by monolayer graphene coverage further away from the steps.16 Thus, for narrower terraces, the areal density of step edges is higher and a relatively larger part of the surfaces is covered by graphene multilayers, resulting in a higher average thickness. The observation that the silane 1 and the standard process yield similar graphene coverages can be explained by the fact that the process temperature was adjusted as already mentioned above. The 125°C higher temperature obviously compensates the reduced density of step edges due to the increased terrace width. Notably, the silane 2 process gave not only a lower graphene coverage as compared to the standard Ar and silane 1 processes for larger miscuts, but also resulted in coverages exhibiting a considerably larger scatter and lacking a distinct trend in the dependence on the miscut angle.
Morphology and local graphene coverage obtained by the silane processes were also investigated by LEEM for selected samples. For graphene on buffer layer on SiC(0001), LEER spectra can serve to unambiguously determine the layer thickness because the number of dips corresponds to the number of graphene layers.33, 34 At a given electron energy, the variation of reflectivity gives rise to different levels of gray in the image. Figure 2(a) shows a LEEM bright field image of a sample surface after the silane 1 process, and typical LEER spectra observed on that sample are given in Figure 2(b). In the micrograph, three different levels of gray can be identified, which can be attributed to mono-, bi- and trilayer graphene. While mostly covered by MLG (bright gray), areas of bi- and trilayer graphene have formed in the vicinity of step edges, which appear as narrow dark lines in the micrograph. The coverage extracted from the LEEM images agrees with the value determined by XPS. The morphology resembles that of SiC(0001) graphitized in argon, which has been investigated by LEEM previously.16 This agrees with the above observation that the graphene coverage follows the same dependence on the miscut angle for both the silane 1 and the Ar processes.
eV) of a sample graphitized in the silane 1 process. (b) Typical LEER spectra observed on that sample. Spectra are offset from each other for clarity. (c, d) LEEM bright field images (
eV and 0.2 eV, respectively) acquired at two locations on a sample subjected to the silane 2 process. The number of graphene layers is indicated in the micrographs.
Figures 2(c) and (d) show LEEM micrographs of a sample prepared by the silane 2 process. On same parts of the surface [Figure 2(c)], at least 
m wide terraces homogeneously covered by monolayer graphene (medium gray) are found. Here, the bilayer stripes (dark gray) are very narrow compared to the terrace width or even absent. On other spots on the same sample [Figure 2(d)], the formation of small bilayer islands on the terraces is observed. Note that due to a different electron energy, in this image bilayer graphene appears in bright gray. The coverage extracted from the LEEM images agrees with the value determined by XPS. Such bilayer graphene inclusions are obviously undesirable for obtaining homogeneous graphene. Furthermore, samples prepared in the silane 2 process often also exhibit areas of ungraphitized SiC (buffer layer and occasionally bare SiC), resulting in average coverages below one monolayer [see Figure 1(f)].
The observed inhomogeneities and the lack of reproducibility of graphene coverage for higher silane flow pose limits for what can be achieved by the addition of silane to Ar in atmospheric pressure graphitization. It is worth mentioning that alternative approaches for obtaining large-area MLG have recently been pursued which rely on avoiding huge step bunching, including the use of 3C-SiC(111) as substrates for graphene growth,46 high-temperature pretreatment,47 polymer-assisted sublimation growth,48 and adjusting the heating rate.49
4 Dark Field Characterization
As discussed in the preceeding section, bright field LEEM gives information about the local coverage with graphene on the surface. However, additional structural information, like, e.g., about stacking of layers, can be derived from LEEM dark field images as demonstrated for graphene on SiC(0001) by Hibino et al.50 and Siegel et al.51 Figure 3(a) shows a LEEM bright field image of SiC(0001) graphitized in the silane 2 process. Two extended monolayer graphene terraces (medium gray) can be seen which are separated by a macrostep in the SiC that appears as an almost vertical line with bilayer graphene (bright gray) in its vicinity. Dark field images (Figure 3(b)–(d)) have been acquired using the satellite diffraction spots around the specular spot. These spots can be described as the result of multiple electron diffraction at the SiC and the graphene/buffer layer.13, 51 Therefore, these dark field images should carry information on both the substrate and the stack on top of it as pointed out previously.51
eV) and (b)–(d) dark field images (
eV) of a mostly monolayer graphene area of a sample from a silane 2 process. The arrows in (a) mark steps in the SiC substrate. The images have been acquired using the specular and satellite spots indicated in the upper insets. The lower insets give an enlarged representation of the areas highlighted by the dotted rectangles.
The first interesting observation in the dark field images is the inversion of the contrast within the monolayer region in the left half of the micrograph when imaging with neighboring diffraction spots [see white dashed line in Figure 3(b)]. This can be explained by the three-fold symmetry of a single atomic terrace of SiC.51-53 On 6H-SiC(0001), there are three surface terminations denoted S1, S2, and S3, which have one, two, or three identically oriented SiC bilayers below the surface, and three surface terminations S1*, S2*, and S3*, which are structurally equivalent to the first but rotated by 60°.52, 53 When there is a step on the SiC surface which separates a terrace with one of the first three terminations from a terrace with one of the latter terminations, the three-fold LEED pattern rotates as well, thus giving rise to the 'termination' contrast observed in Figure 3(b)–(d). An example for such a step would be an SiC bilayer step between an S1- and an S3*-terminated terrace. Note that on the left terrace, a step is visible in the bright field image as a narrow dark line [marked by the white arrow in Figure 3(a)] at the position of contrast inversion in the dark fields [marked by the dashed line in Figure 3(b)]. When a substrate step is not connected with a rotation of the surface termination, no contrast should be visible in the dark field images. This is indeed the case for the step marked by the black arrow in Figure 3(a). The same reasoning can be applied to explain the contrast inversion between the terrace on the right side of the macrostep and that in the lower part of the left side terrace. It should be mentioned that the existence of different SiC terminations is not specific for the silane process.
The second interesting observation is a network of lines that is visible in the dark field images on the monolayer graphene regions. While the direction of the lines is less regular on some parts of the surface, there are also extended regions where the lines run nearly perfectly parallel, for example in the areas marked by the dotted rectangles in Figure 3(b) and (d). A magnified view of these areas is shown as insets in the respective figures. Hibino et al.50 reported LEEM dark field contrast in bilayer graphene residing on the buffer layer on SiC(0001) arising from domains of AB and AC stacking, in accordance with a later study.51 We note however several differences of the present study to these investigations. First, in our case the sizes of the monolayer graphene regions are much larger as compared to the UHV grown samples;50, 51 second, we observe the line contrast on monolayer graphene on the buffer layer; third, the line structures are qualitatively different from the stacking dark field contrast reported in Refs. [50, 51].
On the other hand, the network of lines in our dark field images resembles the results of a dark field transmission electron microscopy (TEM) investigation of freestanding bilayer graphene membranes by Butz et al.54 These membranes have been obtained from monolayer graphene on the buffer layer prepared by argon graphitization of SiC by subsequent photochemical removal of the substrate.55 Butz et al.54 found a line pattern mostly parallel in large parts of the surface with a spacing which is comparable with what we observe in the regions of regular lines [∼50 nm in the highlighted areas of Figure 3(b) and (d)]. They show that their line pattern is due to a dislocation pattern which leads to an alternating AB/AC stacking in the bilayer membranes.54 Furthermore, they observed that the contrast of the dislocation lines depends on the orientation of the diffraction vector (i.e. the diffraction spot used for imaging) with respect to the Burgers vector of the dislocations such that the contrast almost vanished for certain orientations.54 We notice a similar behavior in our LEEM dark field images: Whereas the vertical parallel lines in the dotted region are clearly visible for two of the satellite spots [see insets in Figure 3(b) and (d)], they are not detectable in Figure 3(c). Based on these observations, we propose that the lines in our satellite spot dark field images have the same origin. This would indicate that the dislocations observed by Butz et al.54 in the membranes are present in the system of monolayer graphene on the buffer layer already before the photochemical removal of the substrate and formation of a freestanding bilayer graphene membrane. Furthermore it would confirm their findings on membranes where SiC was only partially removed.54 As a likely origin for the dislocations, the different thermal expansion coefficients of SiC, to which the buffer layer is covalently bound, and the graphene layer is discussed.54 Note, that the dark field line patterns described above are not exclusive to the silane process, but were also observed with graphene prepared by the standard argon process. Recently, the impact of dislocations on charge transport in bilayer graphene obtained by hydrogen intercalation of MLG has been revealed.56 Apparently such fine structural details are of great importance for fully understanding the physics of electronic transport in epitaxial graphene.
5 Growth on Nonpolar SiC Surfaces
As discussed above, epitaxial growth of graphene on SiC has been intensively studied on the hexagonal (0001) and 
surfaces.10 Whereas the above mentioned buffer layer is observed on the (0001) oriented surface, a similar structure is absent on SiC
. The buffer layer on SiC(0001) is responsible for the high intrinsic carrier concentration of the epitaxial graphene21, 22 and for the temperature dependence of the charge carrier mobility.57 On the other hand, growth on the 
surface, in particular the precise control over the number of layers, is challenging and the layers do not exhibit a fixed crystallographic relationship to the substrate. The question remains whether epitaxial graphene can be grown on other surface orientations of SiC and how the properties compare to those of graphene on the basal plane surfaces. Therefore we have investigated the formation of graphene on 4H-SiC
and 
(also referred to as a- and m-planes, respectively) by sublimation growth in Ar.35
The a- and m-plane samples, purchased from Intrinsic Semiconductor, were annealed in the previously described setup37 in 1 bar Ar. Best results were obtained with annealing at 1600°C for 10 minutes, which resulted in a graphene coverage of one to two monolayers as judged from XPS.35
C1s core level spectra of 4H-SiC
and 4H-SiC
surfaces after graphitization are shown in Figures 4(a) and (b), respectively. Two components are visible in the spectra: an asymmetric component at 284.50 eV typical of graphene and the SiC bulk signal at 283.11 eV and 282.95 eV for a- and m-plane, respectively. The position of the graphene C1s level at a somewhat higher binding energy than that of graphite (vertical dashed line in Figure 4) indicates a slight n-type doping. A C1s spectrum of monolayer graphene grown on the polar 6H-SiC(0001) surface is shown for comparison in Figure 4(d). Besides the asymmetric graphene component located at 284.7 eV (indicating stronger n-type doping) and the bulk component at 283.85 eV, two additional components labeled S1 (285.0 eV) and S2 (285.65 eV) are required to reproduce the measured spectrum. The latter are the fingerprint of the buffer layer formed at the interface between epitaxial graphene and SiC(0001).23 Thus, they are also present in C1s spectra of a bare 
layer without graphene on top [see Figure 4(e)]. From the absence of such components on the graphitized a- and m-planes of SiC, it can be inferred that a carbon layer bound to the SiC comparable to the buffer layer is absent on the nonpolar surfaces. This is in agreement with density functional theory calculations.35 Instead, the C1s signals of the graphitized 4H-SiC
and 4H-SiC
surfaces strongly resemble those of quasi-freestanding graphene where the buffer layer has been eliminated by hydrogen intercalation.25, 58, 59 A spectrum of quasi-freestanding monolayer graphene (QFMLG) prepared in this way is given in Figure 4(c) for comparison. We would like to mention that in this case the binding energy of the graphene C1s component is lower than that of graphite due to p-type doping59, 60 and that the varying positions of the SiC bulk signals are a consequence of a different surface band bending.
(a) and 4H-SiC
(b) surfaces. For comparison, spectra of quasi-freestanding monolayer graphene (QFMLG) obtained by hydrogen intercalation (c), regular MLG (d) and 
(e) are also shown. Components labeled SiC originate from the substrate, and asymmetric components G arise from graphene. Components S1 and S2 are related to the buffer layer. The dashed line indicates the binding energy of the C1s signal of graphite. Spectra are offset from each other for clarity. Adapted from Ref. [35].
Information on structural properties of the graphene films was obtained from LEED. Figure 5(b)–(d) depicts diffraction patterns acquired at different spots on the graphitized m-plane sample using an aperture in the LEEM system that limits the probing area on the sample to 
. The electron energy was 46 eV. The spots along the vertical line arise from the SiC substrate, whereas on the perimeter of the images, a number of graphene diffraction spots is visible. In Figures 5(b) and (c), at least five different rotational orientations of the graphene spots can be observed, and in Figure 5(d) two orientations can be seen (note that on this position on the sample, the SiC spots are not visible because the graphene thickness is too large). The micro-LEED images indicate a large degree of rotational disorder in the graphene films prepared on 4H-SiC
, resembling the situation on SiC
[see e.g. Ref. [10] and references therein]. This is in contrast to the observations for graphene grown on 4H-SiC
. Figure 5(a) shows a diffraction pattern of this surface taken at an electron energy of 140 eV with conventional LEED optics (Specs ErLEED). Besides the SiC spots, which form a rectangular pattern, only six graphene spots can be observed. Obviously, there is no significant rotational disorder of the graphene films on the SiC a-plane even on the 
scale probed by macro-LEED.
obtained at 140 eV. The rectangle shows the reciprocal unit mesh of the 4H-SiC
surface. The vectors 
and 
indicate the reciprocal lattice of graphene. The diffraction spots marked by small arrows are due to multiple diffraction. (b)–(d) micro-LEED patterns of graphene on 4H-SiC
taken at 46 eV at different positions on the sample. From Ref. [35].
Graphene coverage and homogeneity were studied on a microscopic scale with LEEM. Following the procedure outlined above, spatially resolved LEER spectra were acquired. Subsequently, false color images were generated by a numerical comparison of the reflectivity curve of each pixel to typical spectra observed on the sample. The results of this procedure are given in Figures 6(a) and (b) for a graphitized m-plane surface. The colors in Figure 6(a) mark the areas with LEER spectra matching the six curves depicted in Figure 6(b). Based on the pronounced minima, the spectra are assigned to one to four monolayers of graphene,33 whereas the flat curve shown in black is attributed to uncovered SiC. The false-color image shows extended areas of mono-, bi- and trilayer graphene on 4H-SiC
after the annealing while at the same time smaller portions of uncovered SiC and four monolayers of graphene are present [Figure 6(a)]. For 2-monolayer (2 ML) regions, two representative spectra were observed which differ mainly in intensity at higher energies. The origin of this behavior is presently not understood.
after graphene growth. The colors indicate matching LEER spectra. (b) Typical LEER spectra of graphitized 4H-SiC
corresponding to the false color presentation. (c) False color image and (d) typical LEER spectra of graphitized 4H-SiC
. The assignment of the spectra to the graphene coverage in monolayers is indicated. The scale bar is 1μm for both images. Spectra are offset from each other for clarity. Adapted from Ref. [35].
The results obtained in an analogous way for 4H-SiC
are displayed in Figures 6(c) and (d). Here the dips in the LEER spectra are less pronounced and appear broader in comparison to the dips observed in the spectra of the graphitized 
surface. Nevertheless, at least in the case of the spectra of one to three monolayers, the number of dips serves as an indicator for the local coverage, as given in Figure 6(d). Note, that two slightly shifted spectra were observed for the 2 ML regions, possibly due to a local variation of the work function.61 From the color-coded image [Figure 6(c)], an average graphene coverage of (2.0 ± 0.1) ML is derived. This is in perfect agreement with the thickness derived from the C1s core level components, thus corroborating the assignment of the spectra. LEEM shows that the graphitization of the a-plane sample resulted in the formation of continuous monolayer graphene terraces with a width of 
m accompanied by bi- and trilayer and to a lesser extent 4 ML graphene patches. In comparison to the m-plane, the graphene thickness on 4H-SiC
is more uniform, possibly qualifying it as an alternative substrate for the sublimation growth of buffer-layer-free graphene on SiC.
With respect to rotational disorder and uniformity graphene on SiC
and 
resemble graphene on SiC
and (0001), respectively. Similar observations were made by STM for the m-plane62 where Moiré patterns were observed. On the other hand, the same group also reported Moiré patterns in some graphene multilayers on the a-plane63 with a small twist angle of 5.4° between adjacent layers as well as Bernal stacked layers. In contrast to our study, they used 6H-SiC, which raises the question about a polytype dependence. This should be topic of future studies.
6 Interface Oxidation in Water Vapor
As mentioned in section 1, buffer-layer-free graphene can be obtained on SiC(0001) when the 
layer is lifted off from the substrate by intercalation. Among the various elements investigated as intercalants, oxygen has attracted the attention of various groups. Different procedures have been employed, including annealing in molecular oxygen at moderate temperatures and high pressure (up to 1 bar)27, 64 and annealing at higher temperatures but at low pressure.64, 65 In an earlier study, we found, however, that a large number of defects is introduced in the decoupled buffer layer for both approaches.64 Interestingly, simple annealing in air has also been reported to decouple the buffer layer and to yield high-quality bilayer graphene when applied to MLG.66 In this investigation, the question was raised whether the precursor for the interface oxidation is the O2 in air or its content of water vapor. To shed light on this, we studied the buffer layer decoupling by interface oxidation employing annealing in a controlled atmosphere of water vapor.36
For these experiments, a petri dish containing deionized water was placed in a vacuum chamber which was evacuated at high speed, causing the water to freeze. After the residual gas in the chamber was removed down to a pressure of 
mbar, pumping was stopped. Subsequent melting of the ice and evaporation created a saturated water atmosphere. Samples were fixed to a conductive heater and annealed for 30 min at 500°C and 650°C for buffer layer and MLG, respectively.
C1s core level spectra of the buffer layer and MLG before and after interface oxidation in water vapor are compiled in Figure 7(a)–(d). As has been described in the previous section, the spectrum of the pristine buffer layer exhibits three symmetric components due to photoemission signals from the SiC and the buffer layer (S1 and S2). The spectrum of pristine MLG shows an additional asymmetric component (G) originating from the graphene layer. In the spectra taken after water treatment, the buffer layer related components are no longer visible. Besides the bulk component at 282.55 eV and 282.9 eV, respectively, only asymmetric components typical of graphene are detectable in the C1s spectra of water-treated 
and MLG samples (with a higher intensity in the latter case). This signals the decoupling of the buffer layer from the SiC substrate and its conversion into a graphene layer, comparable to reports for hydrogen intercalation.25, 59 The position of the graphene components of 284.25 eV for water-treated 
and MLG samples indicates a hole doping of the quasi-freestanding graphene layers.
The buffer layer decoupling is accompanied by an oxidation of the interface as evidenced by the Si2p spectra shown in Figure 7(e) and (f). In the case of the water-treated buffer layer sample, another component is present in addition to the SiC substrate signal which is attributed to the formation of a silicon suboxide (Si+)67, 68 at the interface. The altered interface results in a change of surface band bending witnessed by the different binding energies of the SiC components in the C1s and Si2p core level spectra before and after water treatment. Notably, a small component at a chemical shift typical of Si4 + 67, 68 can be observed for water-treated MLG. This indicates the formation of a thin layer of SiO2 likely related to the higher annealing temperature required for intercalation in the case of MLG. The different interface is believed to be the reason for the slightly different band bending observed for water-treated 
and MLG.
To study the morphology of the samples and the effect of water treatment on a local scale, LEEM was employed. An alignment mark allowed us to investigate the same sample area before and after the interface oxidation. LEEM bright field images of a sample subjected to the standard MLG growth procedure (which resulted in an unusually high graphene coverage in this particular case) before and after interface oxidation are shown in Figure 8(a) and (b), respectively. In both cases four different characteristic LEER spectra were observed. These are labeled A–D for the pristine sample and E–H for the treated sample. They are plotted in Figures 8(c) and (d). Using these typical spectra, color coded images can be generated as described in section 5 which are presented in Figure 8(e) and (f), respectively. Every pixel is color-coded according to the matching standard spectrum. Obviously, areas on the sample which showed reflectivity curves of type A (B; C; D) in the pristine state can be matched to areas characterized by spectra of type E (F; G; H) after interface oxidation. Apparently, there is no change in the surface morphology. Similar observations were made for a water-treated buffer layer sample.36
Hibino et al.33 reported that the number of graphene layers corresponds to the number of dips in the LEER spectra of graphene on SiC(0001). In their scheme, the 
reconstruction, which exhibits a relatively featureless spectrum without a clear dip, is not counted as a graphene layer. Thus, in the pristine sample, areas A can be assigned to the buffer layer, whereas areas B, C, and D with one, two, or three dips in the corresponding spectra are covered by mono-, bi- and trilayer graphene, respectively. In more recent work, LEER spectra of graphene have been re-interpreted. It was concluded that a stack of n graphene layers shows 
reflectivity minima which are related to interlayer states between the graphene layers rather than states located on the graphene sheets.69, 70 Here, the buffer layer is counted as a graphene layer since it has a graphene-like structure.23 For the pristine sample discussed here, both interpretations lead to the same assignment of layer thickness, but this is not the case after water treatment as will be discussed in the following.
With the local coverage identified by the LEER spectra measured on the pristine sample and knowing from XPS that the buffer layer was converted to graphene, one can immediately assign local coverages to the different areas in the water treated sample (except for region H as discussed below). Hence, regions E, F, and G correspond to quasi-freestanding mono-, bi-, and trilayer graphene, respectively. The LEER spectrum of the quasi-freestanding monolayer areas (regions E) exhibits two dips at 1.4 eV and 4.2 eV. Regions of quasi-freestanding bilayer graphene (F) show a dip at 2.8 eV as well as an additional dip at 5.8 eV. Spectra of quasi-freestanding trilayer areas (G) have two dips located at 1.5 eV and 4.0 eV and a shoulder at 5.8 eV. The LEER spectra of quasi-freestanding mono-, bi-, and trilayer graphene obtained after interface oxidation thus indicate, that in this case simple counting of the number of dips following Hibino et al.33 is not sufficient to determine the number of layers. This speaks for Feenstras69, 70 interpretation, who suggested that the spectrum results from interlayer states between the graphene layers in the stack as well as between the substrate and the graphene layer next to it. In this picture, the observation of two dips in the spectrum of quasi-freestanding monolayer graphene (regions E) indicates the formation of two interlayer states between the graphene layer and the SiC substrate probably due to a large separation. For the interpretation of the spectra of regions F (G) the number of dips indicates that the inclusion of one state between the substrate and one (two) state(s) between the graphene sheets suffices. It is worth mentioning that the mere counting of dips, even using Feenstra's re-interpretation, cannot fully capture the underlying physics resulting in the LEER spectra. The shape of the LEER spectra is caused by the unoccupied electronic structure of the sample. This calls for a more elaborate theoretical treatment as alluded to in the review by Flege and Krasovskii.71 Finally, we mention that no significant change in the LEER spectra of the trilayer areas upon annealing in water vapor is observed (spectra D and H). Since these areas only cover a negligible part of the surface, it is impossible to tell from our XPS data if buffer layer was decoupled in these areas.
The decoupling of the buffer layer from the SiC substrate was further confirmed by Raman spectroscopy (for spectra see Ref. [36]), where the presence of the G and 2D peaks after water treatment marks the conversion of the 
reconstruction to graphene. Furthermore, the Raman spectra revealed numerous defects in the converted buffer layer sample but a negligible defect density in the converted monolayer. From Hall effect measurements under ambient conditions, a p-type doping of 
and a charge carrier mobility of 420 cm2/Vs were determined for the converted buffer layer despite the large number of defects. For converted MLG, the mobility was significantly higher with a value of 790 cm2/Vs at a hole concentration of 
. For comparison, quasi-freestanding bilayer graphene obtained by hydrogen intercalation exhibits room temperature mobilities in the range of 
/Vs.59, 72 This suggests that annealing in water vapor is a promising alternative to annealing in hydrogen to produce quasi-freestanding bilayer graphene.
In summary, annealing in water vapor can be employed to decouple the buffer layer similar to annealing in oxygen27, 64, 65 or air.66 Here, too, the SiC surface is oxidized. This implies that the water content in air can play a role in the process carried out in Ref. [66]. Indeed, related work by Bom et al.73 revealed that the combined use of water and oxygen promotes the detachment of the buffer layer from the substrate.
7 Initial Stages of Hydrogen Intercalation
Since the first demonstration of hydrogen intercalation of the buffer layer by annealing in molecular H2 atmosphere by Riedl et al.25 the question about the mechanism has not been addressed experimentally. A process simulation by kinetic Monte Carlo methods, although considering the case of atomic hydrogen, was carried out by Deretzis and La Magna.74 The authors state that their work is also valid for molecular hydrogen when an additional step (dissociative adsorption of hydrogen) is included. Interestingly, the study indicates, that hydrogen is initially adsorbed on the buffer layer. Eventually, the adsorbed hydrogen penetrates the buffer layer and binds to the SiC surface. The sites, where this occurs, are randomly distributed over the surface. With increasing time, the nucleation of clusters of hydrogen bound to SiC is observed, which finally coalesce so that the whole buffer layer is delaminated. This indicates that defects like domain boundaries or surface steps, which are present on real surfaces, are not necessarily required.
In order to test this theoretical prediction, we have initiated a LEEM study of the early stages of hydrogen intercalation. To that end, we have annealed a buffer layer sample on 6H-SiC(0001) in 1 bar molecular hydrogen at a rather low temperature of 470°C for total times of one, three and seven minutes. Following the annealing, the sample was characterized by XPS, LEEM, and LEED. The results are compiled in Figures 9 and 10. The C1s spectrum of the pristine buffer layer (see Figure 9) is composed of the typical components S1 and S2 of the buffer layer accompanied by the signal of the SiC bulk.
After the first minute of annealing in hydrogen almost no changes are observed in the C1s spectrum besides a small shift of the bulk line to lower binding energy, which is caused by a slight change in surface band bending. This might signal the passivation of Si dangling bonds that exist due to the lattice mismatch between the buffer layer and the SiC(0001) surface as was suggested by Pallechi et al.75 After a total annealing time of 3 minutes, the intensity between the buffer layer components and the SiC bulk line has increased. At this position, the peak corresponding to QFMLG is expected. Therefore we can interpret this observation as partial intercalation. At the same time, the bulk line has gained a tail at lower binding energy, which originates in the formation of a new component labeled SiC' that is clearly seen after the next annealing step (7 min). The component SiC' after annealing for 7 min has a binding energy of 282.95 eV which fits well to the position of the bulk line in QFMLG on H-terminated SiC(0001) (see above). The component due to QFMLG is also clearly resolved now. Note that the spectra of the stages of partial intercalation were modeled as a combination of the known spectral features of the pristine buffer layer and fully hydrogen intercalated QFMLG (see Figure 4 and Refs. [23, 60] for comparison), which reduces the number of free parameters in the model. Comparing the intensities of the different SiC components one can conclude that about 50% of the SiC(0001) surface is now covered with hydrogen.
Figure 10 summarizes corresponding bright field LEEM images acquired at an electron energy of 6 eV at which the buffer layer appears darker than MLG. The narrow MLG patches, which cannot be resolved by the XPS measurement, are formed near step edges. After annealing in hydrogen for one minute small bright spots (examples indicated by arrows) appear on the terraces which grow in number and size when the hydrogen annealing is continued (3 min). These spots are thus interpreted as areas of local hydrogen intercalation. While after 3 minutes these patches lead to the signals SiC' and G in the C1s spectrum in Figure 9, their density after 1 minute is too small to lead to corresponding signals in the C1s spectrum that could be unambiguously identified. Interestingly, no evidence for an increased intercalation and buffer layer detachment near step edges is observed. After annealing in H2 for 7 min the contrast between the originally buffer layer covered terraces and the MLG regions at the step edges has reversed indicating the progressing conversion of the buffer layer on the terraces to QFMLG.
Summarizing the described experiments one can state that our preliminary study of the initial stages of H-intercalation under the buffer layer indicates that the process occurs on the terraces and does not necessarily involve step edges. The formation of small intercalated islands was witnessed in qualitative agreement with a Monte Carlo simulation of the process.74 However, additional work is required for a quantitative comparison. On the other hand, the observations discussed above also shed light on the results obtained in the case of interface oxidation via annealing in oxygen, air, or water, where numerous defects were observed for processes carried out on buffer layer samples.36, 64, 66 In these cases, too, it seems that the intercalation process attacks the buffer layer covered terraces directly. If these treatments are applied to MLG, low defect densities were observed. The reason for this different behavior needs to be identified by future work, for which LEEM seems to be an ideal tool as demonstrated here.
8 Summary
The present paper has given an overview of our recent studies of growth and intercalation of epitaxial graphene on SiC, focusing in particular on results from LEEM investigations. Adding silane to the Ar atmosphere in which graphene is obtained by sublimation growth leads to an increased step bunching. The observation of inhomogeneities, e.g. bilayer islands on the terraces as well as uncovered regions, suggests that this method has limited potential for improving the process. Dark field LEEM images taken with superlattice spots from the epitaxial graphene layers indicated the presence of a network of dislocations which were previously observed by TEM investigations of freestanding bilayer graphene membranes that were produced by removing the SiC substrate. The observations indicate that these dislocations are already present before the preparation of the membranes. Their influence on electronic transport should be considered. The growth of graphene layers on non-polar a- and m-planes was also studied and LEEM was employed to investigate the morphology of the films. The studies indicate that the growth on the a- and m-plane is similar to that on the Si- and C-face, respectively. We also employed LEEM to study the decoupling of the buffer layer by interface oxidation in water vapor. By analyzing the same sample area before and after the interface oxidation, we were able to broaden our understanding of LEEM reflectivity spectra. Finally, the initial stages of hydrogen intercalation under the buffer layer was studied. LEEM showed that the hydrogen penetrates the buffer layer on the terraces and that surface steps are not required for intercalation of hydrogen. The results show that LEEM is an excellent tool for studying aspects of growth, structure, and modification of graphene and other two-dimensional materials, which are currently in the focus of numerous researchers.
Acknowledgements
We would like to thank I. Deretzis, F. Giannazzo, G. Nicotra, C. Spinella, and A. La Magna for fruitful collaboration and theoretical support of the work on graphitized nonpolar SiC surfaces. Contributions to sample characterization beyond the scope of this article and support during beamtimes at the synchrotron radiation facility BESSY II by F. Fromm, R.J. Koch, S. Mammadov, P. Wehrfritz, H. Vita, S. Böttcher, and K. Horn is gratefully acknowledged. Furthermore, we thank the group of H.B. Weber for access to their SEM and support of the measurements.
Support of this work by the DFG in the framework of priority program 1459 Graphene is gratefully acknowledged.
Conflict of Interest
The authors have declared no conflict of interest.
