Over the last decade, progress in light-emitting diode performance has been nothing less than phenomenal. LEDs today are performing at 50% wall plug efficiency, meaning that 50% of the applied power is emitted as light. Laboratory results are even higher in the high 60s; these results will become standard in due course. In the meantime, LED-based lighting is replacing incandescent, fluorescent, mercury, and sodium lamps in almost all applications. However, the uptake of LED lighting is still limited by the cost of producing LEDs. This one remaining barrier will be addressed by developments in gallium-nitride-on-silicon (GaN-on-Si) technology.
Shuji Nakamura developed a method of growing thin GaN layers on sapphire substrates in the early 1990s, and up to now these have been the foundation of high-brightness blue LEDs. One notable competitor is silicon carbide (SiC), but these substrates are very expensive. While sapphire costs are dropping, silicon is a very common substrate in the semiconductor industry, and the costs are much lower than either sapphire or SiC.
Semiconductors of all types are characterized by the spacing between atoms in the crystal lattice. One difficulty with using silicon as a substrate is that the atoms are not spaced at the same distance as the atoms in a GaN layer. Growing GaN directly on silicon would lead to a mismatch that would cause strain, and this strain would be relieved only through sporadic dislocations that in turn cause leakage currents and general impairment of the performance of the LED.
The breakthrough needed for growing GaN on silicon was to use a buffer layer that offers a better match to the silicon lattice, and then to gradually transpose the buffer layer into GaN. This buffer technology forms the basis of new GaN-on-Si technology. In addition to the buffer layer, considerable optimization has been pursued