Atomic deuteration of epitaxial many-layer graphene on $4H-SiC(0001¯)$

Chemisorbed hydrogen/deuterium has been shown to modify the electronic structure^1–3 of single-layer graphene as well as induce magnetism.^4–6 It has generated considerable excitement in the scientific community because of its linearly dispersing electronic bands that lead to novel properties, including extremely high electron mobility.⁷ It can be grown epitaxially by sublimation of Si on either the Si-face or the C-face of 4H-SiC.^8–10 Graphene grown on the C-face of 4H-SiC has multiple rotational stacking orders which allows for the growth of many noninteracting graphene sheets¹¹ in contrast to the AB stacking observed on Si-face films.¹² With the very recent discovery of a large bandgap semiconducting form of graphene on SiC substrates,^13,14 there are new possibilities for graphene-based electronics.

The influence of the interface between epitaxial graphene (EG) and SiC has been shown to play a key role in the electronic properties of graphene, namely, on the Si-face where intercalation of hydrogen is shown to create a “quasifreestanding” graphene layer with improved mobility.^15–18 In this case, H₂ is thought to break bonds of Si–C between the substrate and the carbon “buffer,” which otherwise passivates Si dangling bonds from the substrate.¹⁵ Additionally, chemical functionalization and defects in graphene change its behavior, with the potential to create magnetism,^5,6,19 become insulating, modify the wetting properties,²⁰ and change the optical characteristics of the film.²¹ Functionalization and the ability to utilize highly-ordered graphene on substrates for device integration are vital to the future success of graphene-based electronics.

While there has been interest to understand the structure and interface of C-face EG/SiC,²² little work has been done to modify or functionalize it. In order to exploit the possibility of decoupling the high-quality C-face EG/SiC from the substrate, we investigate the use of atomic deuterium as an avenue both to functionalize graphene and to passivate the buried interface.

Epitaxial graphene was grown on C-face 4H-SiC by SiC decomposition using the confinement controlled sublimation (CCS) method described by de Heer et al.²³ 4H-SiC was obtained from Cree, Inc. and cut into 4 × 6 mm² pieces. These samples are sonicated in acetone and isopropyl alcohol prior to growth. After outgassing at 1200 °C, samples are grown inside the CCS graphite crucible with the C-face up at 1450 °C. The CCS method controls the Si vapor pressure during sample growth, allowing for slow sublimation of Si resulting in more uniform samples. The CCS vessel required cleaning for each 15–20 layers of graphene grown on the C-face of SiC for all samples studied of thickness (35L, 60L, and 78L), and this protocol was used to grow the thicker samples. Samples of various thicknesses, 4L, 35L, 60L, and 78L, were used for this study. The thickness is determined by x-ray reflectivity and corroborated by Raman (attenuation) measurements. For studies showing reversibility of atomic deuterium treatment of samples, chemical vapor deposition (CVD) growth of single-layer graphene on copper was used.²⁴ These samples were transferred to SiO₂/Si substrates for Raman measurements.

Atomic deuterium treatment was performed using 1 mTorr of D₂ exposed to a tungsten filament that was heated to 1500 °C in order to crack the molecular deuterium. Treatment was done at room temperature for exposure times of 15 min. In order to minimize damage, the sample prepared for XTEM was treated for 10 min rather than 15 min. Annealing of samples after treatment was done in a vacuum furnace at 1000 °C, where we expect hydrogen to leave the graphene sheet. This temperature is below the mobility of Si and C, where temperatures exceeding 1350 °C are required to begin graphene growth and temperatures exceeding 1200 °C are required to reconstruct the C-face surface of 4H-SiC.⁹ 

Raman spectra were obtained using an Invia Renishaw spectrometer measured at room temperature with a 514 nm laser in a confocal backscattered geometry. The laser power density for these measurements was 2.5 × 10⁴ w/cm² with an exposure time of 100 s. For the measurements on epitaxial graphene on SiC, all intensities were normalized to a bare SiC standard substrate in order to compare intensities of different samples taken at different times. These will be referred to as the “normalized intensity.”

X-ray measurements were performed on a rotating anode with Mo K_α1 radiation utilizing a Ge monochromator to produce a line beam having an angular divergence of 0.003°. Both grazing angle x-ray reflectivity (XRR) and diffraction (XRD) measurements were performed in the specular geometry (wavevector transfer normal to the sample surface). The specular component from XRD is obtained by subtracting a background (off-specular) scan over the same range in q_z where $qz=(4π/λ)sin(2θ/2)$⁠. XRR was taken by integrating the transverse line shape at each q_z, where the specular component was extracted by simultaneous fit of diffuse and Yoneda.^25,26

Cross sections of samples were prepared using lift-out in a focused ion-beam scanning electron microscope (FEI Scios Analytical). Great care has been taken to reduce the Ga ion-beam damage and implantation to the graphene thin film by progressively reducing the beam energies and currents in the cleaning cross-section mode as the sample gets thinner. The sample cross sections were then transferred to a 300 kV transition electron microscope for imaging. The primary electron beam is incident on the cross section perpendicular to the surface direction. Conventional bright field high resolution TEM (HRTEM) images are taken at 300 kV with the sample oriented along the [10-10] zone axis of 4H-SiC. Both samples we investigated are thin and around 25 nm according to the log-ratio method using the zero loss peak and low energy loss peak in the EELS spectrum. The HRTEM images were taken close to the Scherer focus condition, which is about −72 nm. This study was done on a single 60L sample.

Raman scattering gives access to information regarding the carbon hybridization,²⁷ defect density,²⁸ and thickness of graphene samples.²⁹ The “G” (∼1580 cm⁻¹) or graphite peak of the Raman spectra of carbon materials has been correlated with sp² type carbon that should be the only species in pristine graphene. The “D” (∼1350 cm⁻¹) or diamond peak can be correlated with sp³ hybridized defects in graphene.³⁰ 

In order to determine if atomic deuteration would affect our graphene/SiC samples, Raman measurements were first performed on CVD grown single-layer graphene. As shown in Fig. 1, the appearance of the D peak in Raman revealed that chemisorption of deuterium does occur upon atomic treatment and that the process is observed to be reversible, since the D peak disappears by annealing at high temperature.

Epitaxial graphene on SiC, in contrast to single-layer graphene transferred to SiO₂/Si substrates, has additional peaks in the Raman spectra due to SiC.³¹ Raman spectra of a 78L graphene sample are shown in Fig. 2, where the D and G peaks are labeled in order to help distinguish them from the SiC related peaks. The subtracted Raman spectra of graphene are obtained by subtracting the SiC contribution after correction for the attenuation of the beam from graphene.

We analyzed the D peak intensity, which is indicative of the sp³ bonding, in order to look at the effect of deuteration. In Fig. 3(a), the change in the D peak intensity, ΔD, was obtained by subtracting the normalized D peak intensity measured before treatment from the normalized intensity measured after treatment. As can be seen, ΔD increases with the number of atomic layers of graphene, indicating that the number of incorporated deuterium atoms increases with the number of graphene layers. However, the increase with deuteration is rather small, as demonstrated in Fig. 3(b), where ΔD:G is the ratio of the D to G peak normalized intensities measured for the deuterated sample with the same quantity measured before deuteration subtracted. Since the G peak scales with the thickness of the graphene layers, the rapid decrease with thickness in Fig. 3(b) suggests that the attachment of deuterium is not occurring uniformly through the multilayer graphene and it is likely to be attaching closer to the surface and/or buried interface. The latter is suggested because, as will be demonstrated, deuteration strongly modifies the buried interface. The result of Fig. 3(b) clearly indicates that the amount of deuterium incorporating in between the layers is not significant.

In order to investigate the possibility of interlayer incorporation, XRD was used to investigate the graphene interlayer spacing (d-spacing) for the 78L and 35L samples. Figure 4 shows data from the 78L sample, which demonstrates that there is no shift in the first order Bragg position, indicating that there is no expansion or contraction of the graphene d-spacing with exposure to atomic deuterium. Additionally, there are no changes in the shape of the Bragg peaks, as can be seen by comparing an identical solid curve (model obtained from Ref. 26) that is compared to each of the three data sets in Fig. 4. Therefore, atomic deuterium has no apparent effect on the x-ray diffraction.

XRR, which is sensitive to the vertical density profile of the sample but not its crystallinity, was also performed in order to investigate the graphene/SiC interface, as shown in Fig. 5. The data in Figs. 5(a)–5(c) were fit with a reflectivity model³² using standard software³³ that describes the sample composition, layer thickness, and layer roughness for three layers. The layers of the sample are the graphene, the “transition layer” describing the SiC/graphene transition, and the SiC substrate. A schematic of the layer structure used to describe the model is shown in the insets of Figs. 5(d)–5(f), where “T” corresponds to the transition layer. The scattering-length density (SLD) profiles obtained from this analysis, Figs. 5(d)–5(f), for three samples that were atomically treated with deuterium revealed that the interface between graphene and SiC changed dramatically. As can be seen there is a marked SLD decrease, as indicated by the arrows in Figs. 5(d)–5(f). In all cases, the graphene film does not change thickness with treatment, as was shown by x-ray diffraction in Fig. 4.

We interpret the change at the interface to be a signal of the initiation of graphene delamination from the SiC substrate. Given that the total graphene layer thickness does not change, as determined in Fig. 4, we note that the large density decrease at the interface in Figs. 5(d)–5(f) cannot be explained by composition mixtures of graphene and SiC so that the result must be due to the opening of a nanoscale gap (within which we cannot confirm or exclude the presence of deuterium). Indeed, when samples were exposed to two to three cycles of atomic deuteration followed by annealing, the graphene film completely fell off the substrate. We note that this decoupling does not arise from thermal cycling because thick multilayer graphene films on SiC are stable even though they undergo several high temperature (>1400 °C) thermal cycles during the growth process. This suggests that long exposure times to atomic H could be potentially useful for freestanding or (dry) transferrable graphene grown on C-face SiC. While it is known that H (Refs. 15–17) and metals³⁴ can intercalate between few-layer graphene and the Si-face of SiC, our present study using atomic deuterium demonstrates the first signs of complete decoupling of epitaxial graphene from the host substrate on C-face EG/SiC. These results show that XRR is a very sensitive indicator of the initiation of delamination of graphene from the SiC substrate.

The peak observed in Fig. 5(d) for z < 0 in the treated sample is due to a small amount of tungsten that was deposited onto the sample surface from the filament used to crack the D₂ gas. Tungsten was only measurable in this sample as it saw a long exposure time to the filament prior to atomic treatment. An additional layer was added to the reflectivity model, as shown in the inset, from which we estimate ∼20% surface coverage of tungsten islands (based on density) with a thickness of ∼5 nm. The strong x-ray absorption of tungsten causes a slope of the reflectivity below the critical angle, which allowed an accurate determination of this layer without inhibiting the ability to characterize the graphene layer and the buried interface.

In order to further investigate the delamination observed in XRR, XTEM was performed on a sample having 60 atomic layers of graphene. This sample was imaged both before and after exposure to atomic deuterium. Before exposure, Fig. 6(a) shows that the interface between graphene and SiC is sharp and clearly visible on this length scale. However, the addition of deuterium causes a dramatic change at this interface. Figure 6(b) shows the marked change that occurs after exposure to deuterium, evident from the region of lighter contrast and highlighted by the vertical white bars, where a significantly broader interface develops. Although the TEM is unable to identify the nature of this change, it indicates that the exposure to atomic deuterium causes a significant change at the interface. It also corroborates the x-ray reflectivity results that are measured over a much larger (tens of micrometer) field of view than the TEM and are, therefore, able to detect the initial stages of delamination of the epitaxial multilayer graphene from the SiC substrate.

We have shown that atomic deuteration of graphene occurs at the surface, the buried interface, and at defect sites of epitaxial many-layer graphene on C-face SiC. While XRD revealed there was no interlayer incorporation of deuterium, XRR showed that the exposure to atomic deuterium did lead to a significant density reduction at the graphene/SiC buried interface, indicative of delamination of the graphene from the SiC substrate. Multilayer epitaxial graphene films are found to completely delaminate from the substrate after two to three cycles of atomic treatment, indicating a potential pathway to dry transfer of high-quality EG/SiC to other substrates for device integration.

The authors would like to acknowledge the support of National Science Foundation (NSF) under Grant No. DGE-1069091 as well as support from the Oak Ridge National Lab GO! Fellowship. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Workforce Development for Teachers and Scientists, and Office of Science Graduate Student Research (SCGSR) program. The SCGSR program is administered by the Oak Ridge Institute for Science and Education for the DOE under Contract No. DE-SC0014664.