Effects of Trimethylgallium Flow Rate on a -Plane GaN Growth on r -Plane Sapphire during One-Sidewall-Seeded Epitaxial Lateral Overgrowth Hsiao-Chiu Hsu, Yan-Kuin Su 1 , Shyh-Jer Huang, Ricky W. Chuang, Shin-Hao Cheng 1 , and Chiao-Yang Cheng Institute of Microelectronics, Department of Electrical Engineering, and Advanced Optoelectronic Technology Center, National Cheng Kung University, Tainan 701, Taiwan, R.O.C. 1 Department of Electrical Engineering, Kun Shan University of Technology, Tainan 710, Taiwan, R.O.C. Received December 30, 2010; accepted January 25, 2011; published online February 15, 2011 The high crystalline quality of a-plane GaN growth on r -plane sapphire has been demonstrated successfully by using one-sidewall-seeded epitaxial lateral overgrowth (OSELOG). The dislocation density of OSELOG-grown GaN film is 3–4 orders of magnitude lower than that of the as- grown film and the coalescence thickness of OSELOG-grown GaN is less than 5 m. Low temperature cathodoluminescence (CL) shows that the optimum trimethylgallium (TMGa) flow rate during OSELOG plays a significant role in enhancing the crystalline quality of a-plane GaN. The crystalline quality of a-plane GaN can be effectively improved using OSELOG compared with the other ELOG approaches. # 2011 The Japan Society of Applied Physics N onpolar a-plane GaN grown on r-sapphire by metal organic chemical vapor deposition (MOCVD) con- tains a high threading-dislocation density (TDD) of 10 10 cm 2 and a basal-plane stacking-fault (BSF) density of 10 5 cm 1 , due to the planar anisotropic nature of its growth and the large difference in lattice constant between the GaN film and sapphire substrate. Various research groups have devised different methods to overcome the problem, such as epitaxial lateral overgrowth (ELOG) or other modified ELOG techniques, which have been devel- oped to minimize dislocation density. 1,2) Iida et al. 3) have also reported that the one-sidewall-seeded ELOG (OSELOG) could reduce the dislocation density effectively. The advantage of adopting the OSELOG approach can be derived from the high quality a-plane GaN films regrown via one-sidewall of an as-grown GaN seed by tuning the growth- rate ratio of the Ga-face to N-face of GaN films. As a result, the new dislocation is prohibited from propagating away from the coalescence area between the wing and window regions when compared with the conventional ELOG approach. This ELOG technique requires more than 10 m to obtain the full coalescence film thickness with a better coalescence surface, which constitutes a difficulty in layer uniformity control. In this work, a modified OSELOG approach is introduced for improvement in a-plane GaN crystalline quality (TDD  10 6 cm 2 ) and reduction of full coalescence thickness (<5 m). We have observed earlier that the growth rate of the full coalescence step is considered essential for OSELOG GaN film because it influences the strain relation and defect distribution in the regrown film. So far, there is little evidence to substantiate this finding. Before the onset the of OESLOG process, the stripe pattern of a GaN seed on r-plane sapphire was fabricated. The OSELOG approach can be further divided into two steps involving the regrowth of GaN from one-sidewall and the lateral growth mode enhancement for coalescing the GaN film on a SiO 2 mask. In the first step, a seeded GaN layer was grown with a high V/III ratio of 1000. Then, the V/III ratio was subsequently changed to 220 during the GaN growth to obtain fully coalesced GaN on the SiO 2 mask. In fact, reducing the V/III ratio to 220 propelled an increase in the growth rate of the Ga-face sidewall, which resulted in the growth rates of the Ga-face sidewall and the N-face sidewall coming be nearly equal to one another. 4) At the second step, these samples were grown under the same V/III ratio but with TMGa flow rate varied during the coalesced step. The trimethylgallium (TMGa) flow of samples was varied with rates of 15, 30, and 45 sccm. The film surface was fully coalesced and the thickness of these samples was kept identical in order to evaluate the sample crystal quality as a result of varying the TMGa precursor flow rate. For comparison, an a-plane GaN film with a similar thickness was also grown and is henceforth denoted as the as-grown GaN. Figure 1 shows cross-sectional scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of the regrown GaN film after completing the first step, where two regions are divided and shown as a wing region (2 m) and a window region (1.6 m), representing the as-grown GaN seed layer and the regrown GaN film, respectively. The cross-sectional SEM image of the grown GaN layer from one sidewall is shown in Fig. 1(a), where the regrowth of GaN film starts only from the N-face sidewall when the V/III rato is maintained at 1000. It also indicates that the growth rate of GaN film regrown from the Ga-face sidewall is extremely low compared to GaN film regrown from the N-face sidewall. This result seemingly agrees with Iida et al.’s study, 4) demonstrating that the growth rate of GaN film regrown from the Ga- and N-face sidewall could be controlled by changing the V/III ratio. Figure 1(b) shows the cross- sectional TEM imaging of the regrown GaN. It was observed Wing Window 0.5 μm Wing Window 0.5 μm [0001] [0001] [1120] [1100] [0001] [1120] [1100] (a) (b) Fig. 1. Cross-sectional SEM image and TEM image of a-plane GaN after first step of GaN growth with V/III growth rate ratio of 1000.  E-mail address: yksu@mail.ncku.edu.tw Applied Physics Express 4 (2011) 035501 035501-1 # 2011 The Japan Society of Applied Physics DOI: 10.1143/APEX.4.035501