Development of Cost-Effective Native Substrates for Gallium Nitride-Based Optoelectronic Devices via Ammonothermal Growth

Abstract

Since the realization of blue light emitting diodes (LEDs) in 1991 (Nakamura et al., 1991), GaN-based optoelectronic devices have been widely developed and commercialized. Adding blue and green colors to the existing red LEDs enabled full-color displays and white light sources. Blue LEDs were also combined with yellow phosphor to achieve solid-state white light sources. Another major application of GaN-based materials is blue laser diodes (LDs). After the first lasing in 1996 (Nakamura et al., 1996), blue LDs were intensively developed to commercialize high-density digital video disks (HD-DVDs). Starting from its applications to indicators, traffic lights, cell phone back-lighting, and HD-DVDs, GaN-based optoelectronic devices are now used in general lighting and replacing conventional electric light sources using vacuum tubes. Due to the lack of native substrates (e.g., GaN or AlN) in the 1980s and early 1990s, GaNbased devices were developed on heterogeneous crystalline substrates such as c-plane sapphire and c-plane silicon carbide. After the commercialization of free-standing GaN substrates grown by hydride vapor phase epitaxy (HVPE) on gallium arsenide (Motoki et al., 2007), blue LDs have been developed on GaN substrates because highly dislocated devices on heteroepitaxial substrates were susceptible to degradation caused by high current density (1~10 kA/cm2) in laser devices. Nevertheless, people have not switched to GaN substrates for LED applications because of the following reasons: 1) extremely high cost of GaN wafers ($5,000~$10,000 per 2” in 2008) does not fit the cost structure of LEDs; and 2) dislocations do not have deteriorating effects on LEDs owing to the low current density. However, to realize solid-state lighting which replaces existing light bulbs, HBLEDs are required to handle higher current than conventional LEDs. State-of-the-art HBLED chips carry 1 A in 1 mm2 chip (Cree, 2009), which corresponds to 0.1 kA/cm2. Although power saturation at high current injection called “droop” is not an easy problem to solve (Stevenson, 2009), each HB-LED chip will be required to handle even higher current densities to achieve significant reduction of lumen cost in the near future. When the current density becomes comparable to that of LDs, GaN substrate will be indispensable to ensure the reliability of the device. With an increasing demand for low-cost GaN substrates, several