Introduction
The combination of optoelectronic III-nitrides and highly advanced Si technology has the potential to be a key technology for fabricating optoelectronic integrated circuits. The luminescence in the blue and ultraviolet regions of GaN has attracted a lot of interest from many scientists and engineers. It has been reported that cubic GaN (β-GaN) epitaxial films can be grown on GaAs(001) and 3C-SiC(001) substrates by chemical vapor deposition (CVD) [Reference Mizuta, Fujieda, Matsumoto and Kawamura1] or gas-source molecular beam epitaxy (MBE) [Reference Paisley, Sitar, Posthill and Davis2, Reference Okumura, Misawa and Yoshida3]. The successful growth of β-GaN on Si, however, has been reported only by a group from Boston University [Reference Moustakas, Lei and Molnar4]. In a previous paper we reported that thin cubic SiC(β-SiC) formation on Si(001) substrates is effective for the epitaxial growth of β-GaN on Si(001) [Reference Hiroyama and Tamura5]. We confirmed that thin SiC layers not only play the role as a buffer layer to reduce the large lattice mismatch between Si and β-GaN but also as a mask to protect the Si substrates against the irradiation of the active nitrogen, e.g., the nitrogen radicals and atomic nitrogen those were supplied during GaN growth.
The SiC formation by using a carbonization technique on Si substrates has been studied as an initial process of heteroepitaxy of SiC on Si [Reference Pirouz, Ernst and Cheng6]. Carbonization using hydrocarbons is a very easy way to obtain SiC layers on Si substrates; in the conventional CVD the carbonization was performed by using C3H8 [Reference Nishino, Powell and Will7] as a source gas and at very high processing temperatures between 1000 and 1300 °C. On the other hand, it was reported that SiC was obtained at lower temperatures (750 - 1000 °C) using acetylene (C2H2) [Reference Nagasawa and Yamaguchi8], although pit formation in Si substrates was also observed. However low temperature carbonization is desired not only for avoiding the high-density pit formation as much as possible but also for device fabrication. In this paper we report the optimum SiC formation and β-GaN growth conditions, such as V/III ratios and substrate temperatures by RF-activated MBE. We also discuss the quality of β-GaN layers.
Experiment
Hydrogen-terminated Si(001) substrates were placed into a preparation chamber ( having a base pressure of about 5 × 10−8 Torr) which was connected to an MBE growth chamber (base pressure of about 5 × 10−11 Torr). The Si substrates were heated up to 850 to 950 °C for 1 to 10 min under a C2H2 pressure of 5 × 10−7 to 5 × 10−5 Torr in order to form SiC layers. The SiC-formed Si(001) substrates were then cooled down to room temperature and immediately transferred to the MBE growth chamber . They were then annealed at 1000 °C for 1 min and cooled down to a growth temperature of 600 to 900 °C. GaN films were grown using RF-MBE under the following growth conditions: the nitrogen gas flow rate was set at 2 sccm and the RF-power was set at 300 W. Gallium molecular beam was supplied to the substrate by using a Knudsen-cell operating at 850 to 1000 °C.
The surface morphology and the quality of GaN films were characterized by using in-situ reflection high energy electron diffraction (RHEED) and cross-sectional transmission electron microscopy (XTEM) as well as x-ray diffraction (XRD) measurements. Photoluminescence (PL) measurements were also carried out using a 325 nm beam from a He-Cd laser.
Thin Sic Layer Formation
Figure 1 shows high-resolution XTEM (HRXTEM) micrographs of the substrates carbonized at 950 °C for 10 min under various C2H2 pressures. For a C2H2 pressure of 5 × 10−7 Torr small SiC islands were formed on the Si. A RHEED pattern of the as-grown SiC layer showed three-dimensional diffused spots, as seen in the inset of Fig. 1(a). For a C2H2 pressure of 5 × 10-6 Torr the surface morphology of the SiC became flat (Fig. 1(b)), and a streaky RHEED pattern was observed. The film thickness of the SiC was approximately 4 nm. With a higher C2H2 pressure of 5 × 10−5 Torr an approximately 5-nm-thick SiC layer was formed. The clear spotty pattern shown in the inset of the Fig. 1(c) indicates that the surface flatness of the SiC was more degraded than that of the SiC grown at a pressure of 5 × 10−6 Torr. As shown in Fig. 1(b), the spacing between two {111} planes of the SiC layer was approximately 0.25 nm by referring to the spacing of 0.31 nm between {111}Si and was almost the same as that of bulk β-SiC. This indicates that the misfit between SiC and Si was almost completely relaxed by the generation of misfit dislocations in the SiC layer.
Figure 2 shows HRXTEM micrographs of Si substrates carbonized at 950 °C under a C2H2 pressure of 5 × 10−6 Torr for (a) 1 min and (b) 3 min. As seen in Fig. 2(a), a 1-nm-thick SiC layer was formed. The RHEED pattern of the SiC was more diffused than that of the thicker SiC layers. After carbonization for 3 min the thickness of the SiC layer was approximately 3 nm and the surface morphology was two-dimensional (Fig. 2(b)). Further growth of the SiC layer was saturated at 4 nm with the carbonization time up to 10 min as shown in Fig. 1(b). This thickness saturation of the SiC layer is known to be due to the suppression of the out-diffusion of Si atoms from the substrates during the SiC formation [Reference Steckl and Li9]. The GaN films grown on the substrates of Figs. 1(b), 1(c) and 2(b) became single-crystal cubic GaN films, while the films grown on the substrates shown in Figs. 1(a) and 2(a) became polycrystal structures having mixed phases of cubic and hexagonal GaN. Therefore, in order to avoid the degradation of the Si substrate as much as possible during the SiC formation, we employed the SiC formation conditions by short time heating for 3 min under a medium C2H2 pressure of 5 × 10−6 Torr.
General Feature of GaN Growth
Figure 3 shows the GaN film growth rate as a function of Ga-cell temperatures (TGa) and substrate temperatures (Tsub). In the case of a constant Tsub the growth rate increased exponentially with the increase of TGa until it reached a saturation value of 110 nm/h at a TGa of 950 °C. This tendency was observed for all the Tsub cases from 600 to 800 °C. The growth rate at a TGa of 1000 °C was almost the same as that at 950 °C. This shows that the film growth at a TGa below 950 °C was dominated by the amount of Ga-flux (Ga-limited growth) and that the film growth at a TGa higher than 950 °C was dominated by the amount of N-flux (N-limited growth). Furthermore, in the case of a constant TGa below 930 °C the growth rate decreased with the increase of Tsub. This indicates that there was some loss of Ga by desorption during the growth with increased Tsub. With TGa of 950 °C and Tsub of 800 °C we concluded that the growth conditions of GaN were close to the stoichiometric conditions.
Figure 4 shows RHEED patterns of GaN grown with different Tsub and TGa for [110] incidence of electron beams. All the patterns show that the GaN films basically have a β-GaN structure, except for that grown at a Tsub of 900 °C. For a TGa of 870 °C the diffused ring-like pattern indicates that the grown layer consisted of grains that rotated on the (001) plane. For a TGa of 900 °C the ring-like pattern was obtained at a Tsub of 500 °C. With the increase of Tsub the pattern became sharper but showed extra weak spots due to the twin formation at a Tsub of 700 °C. For a higher TGa of 930 °C spotty patterns from β-GaN were observed for all the Tsub. However, in the case of a Tsub of 600 and 800 °C, extra spots were included. At the highest TGa of 950 °C, spotty patterns were slightly elongated longitudinally for Tsub of 700 and 800 °C, which showed that the growing surface as rather flat. However, the spotty patterns at 900 °C mainly originated from a hexagonalstructure together with clear rings due to the polycrystal formation.
Crystal Quality
Figure 5 shows XTEM micrographs of the GaN grown at 600 °C for 1 hour with different TGa. At a TGa of 870 °C a columnar structure was clearly observed due to a large rotation angle of each column and the quality was much poorer than that of the higher TGa case. At a TGa of 900 °C the size of each column was larger than that at 870 °C and boundary became unclear. At a TGa of 950 °C no columnar structure was observed and a nearly flat surface was obtained. With a higher TGa, for example, 1000 °C, the film growth rate did not increase as mentioned before and a Ga droplet formation on the surface became obvious due to the significant amount of excess Ga. These results suggest that the slightly Ga-rich conditions are favorable for the growth of β-GaN on thin-SiC-covered Si(001).
Figure 6 shows XTEM micrographs of GaN films grown with various Tsub. The TGa was kept at 950 °C. In Fig. 6(a) and 6(b) no clear columnar structures were observed, although a number of stacking faults and twin boundaries were formed. At a Tsub of 900 °C, the quality of the GaN film became considerably worse, which coheres with the result of Fig. 4. This may be due to the enhancement of Ga desorption from the growing surface.
Long-time growth of GaN films was then carried out at Tsub of 700 and 800 °C. The TGa was fixed at 950 °C. The RHEED pattern of the 1.1-μm-thick film grown at 700 °C for 10 hours showed almost the same pattern as that after growth for 1 hour. On the other hand, the RHEED pattern of the sample grown at 800 °C for 10 hours showed the appearance of many extra spots, which indicated mixing of a hexagonal phase. From these results we concluded that the optimum Tsub and TGa are respectively 700 and 950 °C.
An XRD spectrum of the thick GaN had only one peak at 2θ = 40.06° translating to the lattice spacing of (002)GaN as 2.25Å, which indicates that the grown GaN film was a single phase of β-GaN. The full width at half maximum (FWHM) of the (002)GaN peak was 27 min. As listed in the ref. [Reference Okumura, Ohta, Feuillet, Balakurishnan, Chichibu, Hamaguchi, Hacke and Yoshida10], the reported FWHMs of β-GaN on GaAs and 3C-SiC was distributed in the range between 16 and 76 min.
Figure 7 shows an 8 K PL spectrum of a 1.1 µm-thick b-GaN film grown under the optimum conditions; TGa and Tsub were at 950 and 700 °C, respectively. The PL spectrum was dominated by a near band-edge peak at 381 nm (3.25 eV). The FWHM of the 381 nm peak was approximately 5.9 nm (130 meV). The FWHM of our b-GaN is wider than those obtained by Okumura et al. for β-GaN on a 3C-SiC substrate using ECR-MBE ( 19 meV at 4.2 K) [Reference Okumura, Ohta, Feuillet, Balakurishnan, Chichibu, Hamaguchi, Hacke and Yoshida10] and by As et al. for β-GaN on a GaAs substrate using RF-MBE ( 24 meV at 2 K) [Reference As, Schmilgus, Wang, Schöttker, Schikora and Lischka11]. The intensity of the yellow band between 500 and 700 nm, which is typically observed for hexagonal GaN, was 1000 times lower than that of the near band-edge peak at 381 nm. The donor-acceptor (D-A) recombination peak, reported to be around 400 nm for MOCVD [Reference Kuzunia, Yang, Chen, Krishnankutty and Ashif Khan12] and gas-source MBE using hydrazine [Reference Okumura, Ohta, Feuillet, Balakurishnan, Chichibu, Hamaguchi, Hacke and Yoshida10] and ammonia [Reference Okumura, Yoshida and Okahisa13], was not observed in our β-GaN.
Conclusions
We investigated the optimum growth conditions for obtaining cubic GaN on SiC-covered Si(001) by using RF-MBE. We found that SiC layers between 2.5 and 4 nm thick became flat and that single-crystal cubic GaN films could be grown on thus formed SiC layers. Under optimized growth conditions it was found, by XTEM observations and XRD measurements, that GaN has a flat surface and the amount of hexagonal GaN is negligibly small. The PL spectrum was dominated by a near band-edge peak at 381 nm whose FWHM was 5.9 nm.
Acknowledgments
The authors thank Dr. Tomoyoshi Mishima and Dr. Seiichi Isomae in Hitachi Central Research Laboratory for PL and XRD measurements. They also thank Dr. Masakazu Ichikawa and all members of the Ichikawa group of the JRCAT for fruitful discussions. This work, partly supported by NEDO, was performed in the JRCAT under the joint research agreement between ATP and NAIR.