20 Years And Counting For The GaN LED
If the readership of Compound Semiconductor were asked to name the inventor of the GaN LED, the most common reply would probably be Shuji Nakamura, the former Nichia researcher who is now an academic at the University of California, Santa Barbara. But this is not the correct answer. The first GaN LED was actually made in 1971 by a team at the laboratories of the Radio Corporation of America (RCA), which included Herb Maruska, Jacques Pankove and Ed Miller.
This device employed a metal-insulator-semiconductor structure, and its incredibly low efficiencies prevented it from ever being a commercial success. The far more important breakthrough is, without doubt, the first LED with a p-n junction, because this is the architecture employed in all of today’s blue, green and white-emitting devices. This LED emanated from the Isamu Akasaki’s group at Nagoka University, Japan. Nakamura contribution was to quickly build on this work by optimizing the design, which led to a hike in device efficiencies and enabled Nichia to launch its first commercial LEDs in 1993.
One of the striking things about Akasaki’s achievement is that it came when he had just turned 60, a time in most researchers’ lives when they have lost their drive for striving for success. But his breakthrough was the culmination of many years of nitride development that stretched back to the early 1970s, when he worked as a research scientist at Matsushita Research Institute Tokyo.
Nitrides were a relatively hot topic back then, thanks to several recent breakthroughs. These included the fabrication of the first single crystals of GaN on sapphire substrates in the late 1960s, which were produced by HVPE. Stimulated emission from optically-pumped GaN followed a few years later, along with the previously mentioned success at the RCA labs.
But progress was short-lived and by the mid-1970s many researchers were turning their backs of GaN and looking to other materials that might produce more fruitful results. The main problems with GaN were its low crystal quality, and very high residual donor concentrations that made it impossible to realize p-type conduction. Controlling the levels of n-type conduction was not easy, either.
These issues did not deter Akasaki, and throughout the early 1970s he focused on the research and development of a GaN LED while working at Matsushita. And some success came in 1975, when he produced single-crystal GaN by MBE. Material quality was not great, but the advance was still large enough to merit a three-year research grant by Japan’s Ministry of International Trade and Industry. This funding for the development of a blue LED based on GaN led to the fabrication of an MIS LED with a world record external efficiency of 0.12 percent in 1978. But these devices were never a commercial success, due to low yields that stemmed from wide variations in the thickness of the intrinsic layer and poor surface uniformity.
Again, Akasaki was not put off by the slow rate of progress, and focused on the positive aspects of the research. Fluorescence microscopy revealed the presence of high-quality microcrystals in tiny parts of the larger crystals, which contained cracks and pits. What’s more, even the cluster of needle like crystals, which Akasaki has described as “GaN fungus", produced very efficient light emission.
These observations led Akasaki to believe that a GaN LED with a p-n junction had great potential for efficient light emission. He thought that it would be possible to produce p-type GaN if epilayers could be grown that exhinited the same crystal quality as the microcrystals. To realize this dream he went back to basics, and focused on crystal growth. In 1979 he made what he has referred to as a “crucial decision" – he switched his growth process from MBE to MOCVD.
His reason for this switch was a belief that MOCVD was a superior growth technology for depositing nitrides on mismatched substrates, thanks in part to minimal reverse reactions. Typical growth rates were ideal for nitride growth, and adjustments to alloy compositions and dopant concentrations could be made by simply varying the flow rates of the gas sources.
In comparison, HVPE was rejected because film growth by this technique suffered from reversible reactions, and the deposition rates were too fast for films with a thickness of just a few nanometers. MBE, in comparison, had the downsides of producing epilayers with nitrogen deficiencies, and operated at low growth rates.
After Akasaki had selected his growth method, he needed to choose a substrate. He selected sapphire, due to its high stability under the growth conditions for GaN - temperatures above 1000 ºC and an NH3 atmosphere. Sapphire also has a crystal symmetry that is similar to that of GaN.
Commercial MOCVD reactors were unavailable in the late 1970s, so Akasaki instructed two of his graduate students, Yasuo Koide and Hiroshi Amano, to build the growth tool. By 1981 they were up and running. However, the initial results were disappointing. Crytsal quality was poor, due to large thermal and lattice mismatches between GaN and sapphire.
To combat this, the team developed a novel low-temperature buffer layer technology. This involved low-temperature growth of an incredibly thin layer of a material with physical properties similar to both GaN and sapphire that led to a high-quality interface, thanks to elimination of interfacial free energy. Candidates for the buffer included AlN, GaN, ZnO, and SiC, and the first of these was selected, due to Akasaki’s familiarity with this material. Success did not follow overnight, but by 1985 this team had grown the world’s first high-quality single crystals of GaN.
The characteristics of these films were incredibly encouraging. Near band edge emission dominated the photoluminescence, and the residual electron concentration was of the order of 1017 cm-3, an improvement of more than two orders of magnitude over previous films. Dislocation density had fallen from more than 1011 cm-3 to 108 - 109 cm-3, and electron mobility rocketed from 20 cm2 V-1 s-1 to 700 cm2 V-1 s-1 All of these improvements were down to the role played by the low-temperature buffer, which provided a high density of nucleation centers with the same orientation as the substrate, and promoted lateral growth of the subsequent epilayers.
P-type doping in GaN was the next challenge. This problem had already attracted the interest of many research groups that were unable to crack it, but Akasaki’s team had the significant advantage of starting with a material containing a far lower residual electron concentration. However, they were unable to realize p-type GaN using a zinc dopant.
An important advance came in 1987, when they discovered that the intensity of zinc-related luminescence increased substantially after this GaN sample, which has been grown with low-temperature buffer technology, was irradiated with electron beams during cathodoluminescence studies. Although these crystals did not show p-type conduction, they paved the way to success. Akasaki’s team switched to developing magnesium-doping in 1988 and the following year they produced high-quality, doped samples with two types of magnesium-based precursors. Irradiating these samples with an electron beam produced low resistivity p-type GaN, and the world’s first GaN LED with a p-n junction followed immediately after.
The efficiency of this device, which emitted in the blue/ultra-violet region, was only 0.1 percent. But this shot up to 1.5 percent by 1992, due to improvements in crystal growth quality. The following year Nichia released a commercial device with 2.7 percent efficiency, and the industry has never looked back since then. Revenues have grown to billions of dollars a year, and penetrated a diverse range of markets that include mobile phone displays and backlights, automobile headlamps, torches, streetlights and TV backlights.
Efficiencies continue to rise, and the next goal is general lighting. This market is being targeted with state-of the-art LEDs that feature sophisticated light extraction technologies, multiple-quantum-well active regions and advanced thermal management packages. However, these devices still tend to share two pieces of DNA with the first GaN LED with a p-n junction – a sapphire substrate; and a low-temperature buffer layer. The advances of Akasaki and his team did not just lead to the making of the first commercially viable LED - they created technologies that have been used to this day for LED manufacture.