Developing the world’s highest output in deep-UV light-emitting diode technology
20 January 2012 (Volume 7 Issue 1)
The RIKEN Advanced Science Institute has made significant advances in deep-UV light-emitting diodes (LEDs) technology for a broad range of new applications, including sterilization and decomposition of environmental pollutants
Hideki Hirayama
Team Leader
Terahertz Quantum Device Team , RIKEN Advanced Science Institute
Deputy Team Leader
Deep UV-LED Laboratory, RIKEN Innovation Center
Light-emitting diodes (LEDs) are small, low-energy light sources with a long service life. The LEDs emit infrared rays, visible light, or ultraviolet (UV) rays. In particular, light in the deep-UV band with short wavelengths of between 220 and 350 nanometers has a high sterilizing power that has growing potential in a broad range of applications, including in medicine and in the rapid decomposition of environmental pollutants, such as dioxins. However, deep-UV LEDs(DUV-LEDs) are slow to be adopted because the deep-UV light source is bulky, short-lived, and expensive. Hideki Hirayama, team leader of the Terahertz Quantum Device Team, RIKEN Advanced Science Institute, and Deputy Team Leader of the Deep UV-LED Laboratory, has launched major breakthroughs toward practical applications of this developing technology.
Deep-UV LEDs expected to establish a market of several hundred billion yen
A light-emitting diode (LED) is a light source that enables the production of devices with low energy consumption, long service life, and decreased size and weight. An LED comprises an n-type semiconductor, in which negative charged electrons are the majority carrier, and a p-type semiconductor, in which positively charged “holes” are the majority carrier (see Fig. 1). When a negative voltage is applied to the n-side, and a positive voltage is applied to the p- side, the holes and electrons move and an electric current begins to flow. Upon collision and binding of the holes and electrons in the light-emitting region, they lose their original energy, and the resulting excess energy turns into light energy, thereby producing emission.
Figure 1: Basic structure of a deep-UV light-emitting diode (LED)
The structure of an LED comprises an n-type semiconductor, in which electrons are the majority carrier, and a p-type semiconductor, in which holes are the majority carrier. When a negative voltage is applied to the n-side and a positive voltage is applied to the p-side, holes and electrons move and an electric current begins to flow. Upon collision and binding of the holes and electrons during migration, they lose their original energy, and the resulting excess energy turns into light energy and produces emission.
In recent years LEDs have been developed that emit light over a broad range of wavelengths, including blue and green colors. However, wavelengths in frontier bands, that is, deep-UV radiation with wavelengths from 220 to 350 nm, remain to be further developed. The only currently available methods for generating deep-UV radiation include those using an ultraviolet laser or gas lamp; however, these approaches produce weak emission, require a large, short-lived light source, and are expensive and impractical. For these reasons, there is strong demand for developing a practically applicable LED that emits deep UV.
“This technology has the potential for very wide applications,” Hirayama explains. “For example, a deep-UV radiation with a wavelength of 270 nm can not only be used in medical settings, but also in small sterilizing lamps in refrigerators, air purifiers, and other appliances in ordinary households. Deep-UV radiation with a wavelength of less than 250 nm can be utilized in a high-density recording laser that can store three to four times more data than currently available optical discs. This capacity enables three to four movies of high image quality to be recorded on one disc. Deep-UV radiation with wavelengths of 270 to 320 nm can be utilized for the treatment of environmental pollutants which are difficult to breakdown naturally, such as dioxins, and those with a wavelength of 340 nm can be utilized for highly brilliant white light lamps instead of fluorescent lamps. The market scale for DUV-LEDs for sterilization- related applications alone is estimated at several hundred billion yen per year.”
External quantum efficiency improved by two digits
“An LED is a luminescent semiconductor device, and the wavelength of the light emitted varies depending on the material,” says Hirayama. “Aluminum gallium nitride (AlGaN) is recognized as the best material with which to prepare a DUV-LED. AlGaN is a crystalline alloy of aluminum nitride (AlN) and gallium nitride (GaN), and by changing the content ratio of aluminum and gallium, it is possible to produce light in a broad range of wavelengths, from 200 to 360 nm, which includes deep-UV radiation. AlGaN has many other advantages, such as its high durability and long service life, its applicability to both the p-type and the n-type, and the fact that it is environmentally harmless because it is free from arsenic, lead, and mercury.”
Engaged in DUV-LED research and development for nearly 15 years, Hirayama has seen LED development come a long way since 1997. “In those days, the external quantum efficiency of AlGaN DUV-LED was about 0.01%, a level falling far short of practical application. Raising the efficiency would lead to the emission of light of a higher energy. External quantum efficiency is the ratio of light energy that can be obtained outside the device relative to the electrical energy input, and this is used as an index of an LED’s luminescence efficiency. External quantum efficiency is calculated by multiplying the electron injection efficiency by the internal quantum efficiency and the light extraction efficiency.” Electron injection efficiency indicates the ratio of electrons injected into the emitting layer to all those electrons involved. Internal quantum efficiency is the ratio of electrons that have bound themselves to holes and produced light energy, relative to all the electrons injected into the emitting layer. Light extraction efficiency shows the ratio of light energy obtained outside the LED relative to light energy produced by the emitting layer.
“The level of external quantum efficiency that I have just mentioned, namely about 0.01%, is obtained from electron injection efficiency (about 20% ≒ 0.2) by internal quantum efficiency (less than 1% < 0.01) by light extraction efficiency (about 8% ≒ 0.08). The electron injection efficiency of the AlGaN DUV-LED that we developed in 2010 is up to 80%, its internal quantum efficiency is 50%, and its light extraction efficiency is 8%. The external quantum efficiency is now about 3%,” explains Hirayama (see Fig. 2). These recent developments provide a stark contrast to 1997, when the output from a DUV-LED developed with AlGaN was so low at 20 microwatts that the LED barely emitted any light at all.
New methods in crystal growth increase internal quantum efficiency to 50–80%
Several layers created by the crystallization process are key structural components of an AlGaN DUV-LED. “Generally, numerous semiconductor functions are exhibited by a number of semiconductor layers with different profiles that are several nanometers to micrometers thick, stacked on the surface of a high- quality substrate crystal. An AlGaN DUV-LED also has a similar multilayer structure. A gasified material is sprayed over a sapphire substrate and thermally decomposed to cause crystallization (during vapor phase growth) and creates several important layers in the following order: a buffer layer, an n-type semiconductor layer, an emitting layer, a p-type semiconductor layer, and a contact layer,” says Hirayama (see Fig. 3).
Figure 3: The basic structure of AlGaN deep-UV LED This is the basic structure of an AlGaN deep-UV LED developed in 2010. Internal quantum efficiencies of 50–80% were realized through some key technical breakthroughs, including improved crystal quality for the AlN buffer layer and introduction of an In-incorporated quantum well into the emitting layer. Introduction of a multiquantum barrier (MQB) into the p-type semiconductor layer resulted in an electron injection efficiency of 80%. Shown above is a CCD camera photograph of how the LED is emitting light. A wavelength of 250 nm with an output of 15 mW was achieved, clearing the requirement for practical use in sterilizing lamps. Although ultraviolet light is invisible, this emission looks bluish white because visible light with longer wavelengths is simultaneously generated at an extremely low intensity. The ultraviolet light shines at an intensity several hundred times greater than the visible light.
enlarge image
The quality of AlGaN crystals contributes to a low internal quantum efficiency of AlGaN DUV-LED. High-quality crystals are characterized by (1) a low density of penetration dislocations (discontinuity of atomic arrangement during crystal growth reaching the crystal surface), (2) the absence of cracks, and (3) the maintenance of crystal surface flatness at the level of one atomic layer. Therefore, a low-quality crystal that does not meet the requirements of (1) to (3) is problematic in that the prevalence of hole-electron bonding without emission (non-radiative recombination) increases, resulting in considerably reduced internal quantum efficiency.
“This problem arises from the difference in the distance between the sapphire substrate and AlGaN in the crystal lattice,” explains Hiyarama. “Because they have different lattice distances, their atoms do not become well stacked together, even when forced. If AlGaN crystals are allowed to grow on the sapphire substrate in this state, the penetration dislocation density increases and cracking is likely, so that only low-quality crystals are obtained. As a solution, we interpose an AlN buffer layer to modify the difference in lattice distance, but even this approach does not ensure satisfactory quality. In 2007, we succeeded in solving this problem by developing a new method called the ‘ammonia pulse-feed multilayer growth method.’”
This groundbreaking method enables the growth of high-quality crystals (see Fig. 4). “Traditionally, an AlN buffer layer is created on a sapphire substrate by allowing AlN crystals to grow while gaseous aluminum and gaseous ammonia are fed simultaneously and continuously. In our newly developed technique, we use alternative combinations of two methods of crystal growth. Namely, the conventional continuous-feed rapid-longitudinal growth mode, and the pulse-feed laterally enhanced growth mode, in which gaseous aluminum is continuously supplied and gaseous ammonia is fed in pulses.”
First, nuclei for high-quality AlN crystals are created on a sapphire substrate (see Fig. 4 (1)). Next, the crystals are grown laterally using the pulse-feed laterally-enhanced growth mode to fill the internuclear gaps (Fig. 4 (2)). This process significantly reduces the penetration dislocation density to meet the first requirement. Next, the crystals are rapidly grown longitudinally using the continuous-feed rapid-longitudinal growth mode (Fig. 4 (3)). In this process, the second and third requirements are fulfilled, that is, the maintenance of surface flatness at the level of one atomic layer, and the prevention of cracking. Subsequently, both modes are alternately applied to enable further crystal growth (Fig. 4 (4)).
Figure 4: Improving the quality of AlN buffer layer using the ammonia pulse-feed multilayer growth mode
(1) Nuclei of high-quality AlN crystals are generated on a sapphire substrate. (2) The crystals are grown laterally using the pulse-feed laterally-enhanced growth mode to fill the gaps between the nuclei. (3) The crystals are grown rapidly longitudinally using the continuous-feed rapid-longitudinal growth mode. (4) The pulse feed laterally-enhanced growth mode and the continuous-feed rapid-longitudinal growth mode are alternately repeated to grow the crystals in a multilayer structure. This series of steps produces a low penetration dislocation.
Hirayama adds, “We were able to reduce the penetration dislocation density of the AlN buffer layer to a level one-eightieth of the conventional level by using the ammonia pulse-feed multilayer growth mode. When we stacked an n-type semiconductor layer, an emitting layer, and other layers onto the AlN buffer layer, we could obtain light with a 260 nm wavelength and an output intensity of about 2 mW. The internal quantum efficiency increased dramatically to 50 times higher than the conventional level.”
In addition to these layers, Hirayama and his colleagues introduced an extra three, very thin layers which create a ‘quantum well’ effect of electron enclosure in the emitting layer, whereby the layer encloses both electrons and holes to facilitate their bonding (see Fig. 3). In 2008, the team succeeded in further improving the internal quantum efficiency by introducing indium (In) to AlGaN, and increasing the hole concentration in a p-type semiconductor layer, thus developing an InAlGaN DUV-LED in the 280 nm wavelength band, which is indispensable to the sterilization of objects and the treatment of environmental pollutants. The LED realized an internal quantum efficiency of 80% and an output of 10 mW. These achievements resulted in a significant improvement of the internal quantum efficiency from less than 1% to between 50–80%.
Multiquantum barrier increases electron injection efficiency to 80%
In a world-first attempt, Hirayama and his colleagues introduced an electron-blocking multiquantum barrier (MQB) into an AlGaN DUV-LED (see Fig. 3). Later in 2010, they achieved a dramatic increase in electron injection efficiency to more than 80%, succeeding in generating light at a wavelength of 250 nm with a high output of 15 mW.
According to Hirayama, “The low electron injection efficiency was due to a large percentage of electrons injected into the emitting layer. These electrons passed through it and reached the p- type semiconductor layer. Essentially, the electrons injected into the emitting layer should bind to positively charged holes and emit light. If they pass right through the emitting layer, the injected electrons in effect are wasted. To counter this problem, we introduced an electron-blocking MQB into the p-type semiconductor layer in an attempt to reflect the electrons entering the MQB and return them to the emitting layer.”
Hirayama and his colleagues first tested a single- layer barrier. Electrons with an energy level lower than that of the barrier were reflected back to the emitting layer, but those with higher energy failed to be reflected and reached the p-type semiconductor layer through the emitting layer. “So we then attempted to apply the MQB to obtain multiple barriers. We found that, due to a quantum physical multiple reflection effect, even electrons with an energy level higher than that of the barriers were reflected back to the emitting layer,” says Hirayama. An examination of the MQB using transmission electron microscopy reveals a six-layer structure, with the layer becoming increasingly thicker toward the emitting layer side and thinner toward the p-type semiconductor layer side (see Fig. 3). Of the electrons entering the MQB, those with energy levels lower than those of the MQB are reflected by the first thick layer. The electrons with higher energy levels pass the first layer but are reflected by the second and subsequent layers. “The electron block height of an MQB can be increased to a maximum that is about double that of a single barrier,” adds Hirayama.
The introduction of the MQB approach led to a dramatic improvement in electron injection efficiency from less than 20% to over 80%, and the external quantum efficiency increased from 0.4% to 1.5%. The output reached 15 mW in continuous operations at room temperature. These figures currently represent the world’s highest levels.
“The major feature of an MQB is that even when a large current is supplied, the barrier effect of reflecting electrons and returning them to the emitting layer does not weaken,” explains Hirayama. “In other words, even when the current increases, the efficiency does not decline. This is important in considering the introduction of semiconductor devices into practical applications.” This MQB theory is also applicable to LEDs other than DUV-LEDs. In blue LEDs, reductions in output with an increase in the current have been problematic. Use of an MQB enables a sufficient output to be maintained even with large currents, so the theory is likely to be applicable to LEDs for lighting lamps and other devices that require a certain level of output. Another favorable feature of an MQB is that it can be created using existing equipment for LED production.
Aiming for 70% light extraction efficiency
The key challenge remaining for Hirayama’s team is improving the rate of light extraction efficiency. Currently the light that travels just below the emitting layer can be extracted from the light issued by the emitting layer, however, the remaining light that is scattered to the other layers is wasted, so that light extraction efficiency only reaches a maximum of 8%.
“Since scattered light is mostly absorbed in the contact layer on the p-type semiconductor layer, we thinned the contact layer. We are also investigating a method in which an aluminum-based electrode having a reflectivity of more than 80% is used as the p- electrode in place of a nickel-gold electrode, with a reflectivity of about 20%. We are also looking at an approach to forming a photonic crystal on the sapphire substrate to increase the amount of light extracted. Currently, the light extraction efficiency is 8–15%, but it seems possible to raise this level to 40% by applying these techniques in combination,” says Hirayama. He is also promoting research into a new structure for light extraction to be introduced into the p-type semiconductor layer. Its success would make an efficiency of 70% a reality.
“For DUV-LEDs, we have reached the levels of an external quantum efficiency of 3.8% and an output of 30 mW. However, the external quantum efficiency of a blue LED is more than 80%. The external quantum efficiency of a DUV-LED would not exceed 80% unless the efficiency is 90% or more for electron injection efficiency, internal quantum efficiency, and light extraction efficiency. I want to approach this level for DUV-LEDs. The only problem to be resolved is how to improve the light extraction efficiency. I aim to clear the immediate goal of 70% as soon as possible and bring a DUV-LED into practical application for a broad range of areas.”
Aiming to achieve world-class performance for both the shortest and longest wavelength
In addition to his research into deep-UV light, Hirayama is also investigating terahertz light (wavelength range: 3 mm – 30 μm). With its binary aspects of electronic waves (permeability) and light (light collection resolution), terahertz light is expected to find applications for the internal examination of a wide variety of objects. Although deep-UV radiation and terahertz light differ 1,000 times in wavelength, Hirayama hopes to achieve the world’s number one performance for both of them. “I am so interested in technical development that I can’t help being involved,” says Hirayama.
About the Researcher
Hideki Hirayama
Hideki Hirayama was born in Ibaraki, Japan, in 1966. He obtained his PhD of Engineering from the Tokyo Institute of Technology in 1994. In the same year, he became a research scientist at RIKEN. In 2005, he progressed to the position of team leader of the Terahertz Quantum Device Team, serving concurrently from 2004 as visiting assistant professor at Saitama University. In 2009, he was appointed as visiting professor of Saitama University. He won a Young Scientist Award from the Minister of Education, Culture, Science and Technology in 2005, a Japan IBM Science Prize in Electronics in 2010, and an Ichimura Science Prize in 2011. His research focuses on crystal growth of AlGaN based nitride-semiconductors and development of deep-ultraviolet (DUV) light-emitting diodes (DUV LEDs) and laser diodes (LDs). He is also developing terahertz quantum cascade lasers (THz-QCLs) and terahertz sensing devices based on intersubband optical transition of semiconductor quantum cascade structures.
