Radio-frequency plasma-excited molecular beam epitaxy growth of GaN on graphene/Si(100) substrates

III–nitride semiconductors, including GaN, AlN, InN, and their alloys, have attracted considerable attention because of their outstanding material properties leading to many applications for light-emitting diodes (LEDs), laser diodes (LDs), and high-frequency and high-power transistors. Although many types of substrate have been tried for the hetero-epitaxial growth of III–nitride semiconductors, Si is considered one of the most attractive substrates owing to its high quality at a low cost, ease of obtaining large substrates, and conductivity control. Thus far, c-plane (0001) GaN-based devices such as LEDs and field-effect transistors (FETs) grown on Si(111) substrates have already been fabricated and commercialized. To realize the integration between mature Si-based electronic devices on Si(100) and GaN-based optoelectronic devices, c-plane (0001) GaN growth on Si(100) substrates is essential. However, owing to the different symmetries in hexagonal (0001) GaN and cubic (100) Si, it is still difficult to grow high-quality epitaxial GaN films.^1–4) Therefore, several attempts to integrate GaN FET grown on Si(111) and Si(100) MOSFETs by mechanical bonding have been carried out.^5–7) Recently, it has been reported that graphene can be used as an intermediate layer to grow c-plane GaN on amorphous substrates.^8–10) Since graphene has a hexagonal sixfold symmetry, it should be effective for forming an epitaxial relationship with the c-plane of GaN, even on substrates without a hexagonal sixfold symmetry. In this study, for the first time, we report on GaN growth on graphene/Si(100) by radio-frequency plasma-excited molecular beam epitaxy (RF-MBE).

We used graphene transferred onto Si(100) as a substrate. First, graphene was grown on copper foil by chemical vapor deposition using methane as the carbon source. Then, graphene was coated with a layer of polymer, poly(methyl methacrylate) (PMMA), in order to protect it during the etching of the copper using ferric chloride-containing solutions. We then placed it on the Si(100) substrate. In the final step, the polymer layer was removed using an organic solvent, such as acetone. The graphene layer has a thickness of 1 ML and a multidomain structure (polycrystalline) with different grain sizes. GaN was then grown directly on the graphene/Si(100) substrates by RF-MBE (EpiQuest RC2100NR). Active nitrogen was provided by a nitrogen plasma source (SVT Associates 6.03). After thermal cleaning of the substrate in a vacuum chamber, GaN growth was carried out at 770 °C. The RF plasma power and nitrogen flow rate were 200 W and 2.0 sccm, respectively. The film thickness was about 400 nm (1 h of growth). For comparison, GaN was also grown under the same conditions on Si(100) without graphene. The structure of GaN was monitored in situ by reflection high-energy electron diffraction (RHEED) analysis. After the growth, the films were investigated by scanning electron microscopy (SEM; Hitachi S4300SE) and high-resolution X-ray diffraction (HRXRD) analysis (PANalytical X'Pert MRD). Optical properties of the samples were characterized using cathodoluminescence (CL) at room temperature (GATAN MonoCL2 system installed on Hitachi S-4300SE). The microstructure of GaN was also investigated by transmission electron microscopy (TEM) at 200 kV (JEOL 2010).

Figures 1(a) and 1(b) show plan-view and cross-sectional SEM images of GaN grown on Si(100), with and without graphene, respectively. We found that both GaN films had a columnar structure. As shown in Fig. 1(a), relatively large grains of GaN of 0.5–1 µm size were observed in GaN grown on graphene/Si(100). These large grains clearly show a hexagonal symmetry, indicating c-axis-oriented growth. As shown in Fig. 1(b), GaN grown without graphene on Si(100) had a dense and small grain structure. The size of these small grains was about 100–200 nm.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The crystal structure and growth orientation of GaN films were characterized by HRXRD. Figure 2 shows an XRD profile of ω–2θ measurement of GaN grown on graphene/Si(100). Clear diffraction peaks of (0002) hexagonal wurtzite GaN were observed at ∼34.6°. This indicates that GaN grown on graphene/Si(100) has a c-axis-oriented hexagonal wurtzite structure, in good agreement with the SEM results. In addition to the diffraction peaks of the c-plane GaN, there is a weak peak at 32.4°, which was assigned to the ($10\bar{1}0$) of the m-plane GaN. These results suggest that there is a very small inclusion of grains with different growth orientations in the c-axis-oriented GaN grown on graphene/Si(100), as also seen in Fig. 1(a). We should note that although a (0002) GaN peak was observed from GaN grown without graphene on Si(100) as well, the peak intensity was twenty times lower than that of GaN grown on graphene/Si(100).

Zoom In Zoom Out Reset image size

Figure 3 shows ω-scan (rocking curve) profiles of (0002) GaN grown on Si(100) with (red line) and without (black line) graphene. The full width at half maximum (FWHM) of the (0002) rocking curve of GaN is directly related to the distribution of growth orientation. The FWHM of GaN grown on graphene/Si(100) was 11.3 arcmin. This value is not as good as that of the conventional (0001) GaN grown on sapphire (0001), but lower than those of (0001) GaN grown on Si(100) by metal organic vapor phase epitaxy using an AlN seed layer (16.1 arcmin)⁴⁾ and (0001) GaN grown on ZnO/graphene/silica glass (48 arcmin).⁸⁾ It should also be noted that from GaN grown on Si(100) without graphene, a (0002) rocking curve peak was not observed, as shown in Fig. 3, which means that the c-axis orientation distribution (tilt distribution) was very large for this film. These results confirmed that the insertion of graphene was very effective for growing c-axis-oriented hexagonal GaN on Si(100) substrates.

Zoom In Zoom Out Reset image size

The optical properties of the samples were investigated by CL with an electron beam acceleration energy of 5 kV at RT. Figures 4(a) and 4(b) show SEM and panchromatic CL mapping images of GaN grown on graphene/Si(100). Each grain demonstrated clear luminescence. In particular, strong luminescence was observed from large grains, as shown in Fig. 4(b), which have a hexagonal c-axis-oriented shape. It is suggested that these grains have a higher crystallinity than misoriented grains. Figure 4(c) shows the CL spectrum measured from these bright regions. A CL peak near the band-edge emission energy of approximately 3.4 eV was clearly observed. However, another peak was also observed at approximately 3.2 eV. This luminescence seems to originate from the cubic phase of GaN. It should be noticed that no clear luminescence was observed from GaN grown on Si(100) without graphene.

Zoom In Zoom Out Reset image size

Microstructures of GaN films were investigated by TEM. Figures 5(a) and 5(b) show cross-sectional bright-field TEM images of GaN grown on Si(100) without and with graphene, respectively. The insets are selected area diffraction (SAD) patterns taken from the observed region in Fig. 5. As shown in Fig. 5(a), small grains with widely distributed growth orientation were dominant for GaN grown on Si(100) without graphene. The SAD pattern of these grains also demonstrated a mixture of randomly oriented patterns. In addition, several defects that originated from the high-density grain boundaries in the initial growth region are revealed in the TEM image. Figure 5(b) shows a typical large single domain of hexagonal (0001) GaN grown on the graphene/Si(100) substrate. The SAD pattern of the GaN domain shows a single diffraction pattern of the (0001) c-axis-oriented Wurtzite structure. The epitaxial relationship between the GaN domain and the Si(100) substrate shown in Fig. 5(b) is (0001)_GaN ∥ (100)_Si and $[11\bar{2}0]_{\text{GaN}}\parallel [011]_{\text{Si}}$. Note that there are other c-axis-oriented GaN domains with different in-plane epitaxial relationships observed by cross-sectional TEM. In the GaN grain, there is no threading dislocation, possibly owing to strain relaxation at the grain surface or effective lattice constant modification at the GaN/graphene interface.¹¹⁾ On the other hand, high-density stacking faults were generated in the GaN grain. These stacking faults could be an origin of the inclusion of the cubic-phase GaN. We carried out high-resolution TEM observations in the interface region between GaN and the graphene/Si(100) substrate. As shown in Fig. 6, c-axis-oriented hexagonal (0001) GaN was grown directly on the graphene layer formed on the Si(100) substrate. In the initial 10 nm of growth, however, a high density of stacking faults was generated. As a result, a very thin cubic phase of GaN was grown in the defective initial region. Even after hexagonal GaN growth restarted in the upper region, stacking faults were generated repeatedly. Therefore, such cubic-phase GaN inclusion and a high density of stacking faults should be the origin of luminescence at 3.2 eV.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

These results indicate that the insertion of a graphene layer is very effective for growing c-axis-oriented hexagonal GaN on Si(100) substrates. As shown in Figs. 1 and 5, however, GaN grown on graphene/Si(100) demonstrated grain structures, and their in-plane orientation was not aligned. Although theoretical calculations suggested a stable GaN–graphene interface with a specific epitaxial relationship,^11,12) it is true that the lattice mismatch between GaN and graphene with a 30° rotation is very large (∼30%). Hence, it is easy for GaN to grow with various epitaxial relationships on graphene, as well as in the case of direct InN growth on sapphire.¹³⁾ Nitridation of sapphire substrates by nitrogen plasma is a very important way of fixing the epitaxial relationship between InN and sapphire. Therefore, it is considered that effective surface modification such as nitridation is necessary to obtain a large area of single GaN domain on graphene/Si(100), which is essential for device fabrication. In addition, it is also found that stacking faults were generated at the initial growth of GaN on graphene/Si(100), which resulted in the cubic phase inclusion. Owing to the lack of chemical reactivity of the atomically flat graphene surface, it seems to be also challenging to control the stacking sequences of GaN growth on graphene. It is well known that the step-controlled epitaxy using off-axis SiC{0001} substrates can totally suppress the inclusion of SiC polytypes to grow high-quality epitaxial SiC films.¹⁴⁾ Therefore, the introduction of a step structure on the graphene/Si(100) substrate might be effective for suppressing the generation of stacking faults by step-flow growth.

In summary, we have grown GaN on graphene/Si(100) by RF-MBE. From the characterization by SEM and XRD, we confirmed that GaN grown on graphene/Si(100) shows a c-axis well-oriented hexagonal structure. The FWHM of the (0002) XRD rocking curve of GaN grown on graphene/Si(100) was 11.3 arcmin. On the other hand, we did not observe a (0002) rocking curve peak of GaN grown on Si(100) without graphene. The CL spectrum had a peak near the band-edge emission energy of approximately 3.4 eV. However, another peak was also observed at approximately 3.2 eV, which originated from the cubic phase of GaN. By high-resolution lattice image observation, we found that the growth of the c-axis-oriented GaN started directly on the graphene layer formed on Si(100). However, stacking faults were observed through the GaN grain, especially in the initial growth region. We also observed a very thin cubic (111) GaN inclusion in the GaN columnar grain, which originated from the stacking fault.

Acknowledgments