Thin film LEDs gaining ground

Thin film LEDs gaining ground Excimer laser lift-off enables high brightness LED production Excimer laser-based Laser Lift-Off (LLO) is a key enabling technology in fabricating thin film LEDs. This article reviews the market drivers for high brightness LEDs, current LED designs and their limitations and the basics of thin film devices as well as the LLO process for manufacturing them. The market for high brightness light emitting diodes (LEDs) is growing rapidly, but market penetration is being slowed somewhat by the high price of retrofit LED lamps. One reason for this high price is the drop in efficiency of LEDs at high current densities which increases the number of LED chips necessary to equal the brightness of a typical incandescent lamp. The other reason is that current fabrication technology is still relatively costly. Both limitations can be overcome by a move from the traditional LED device structure to so called thin film devices. Ralph Delmdahl Dr. Ralph Delmdahl is Product Marketing Manager at Coherent GmbH in Göttingen, Germany. His current responsibilities include providing strategic directions for the company s excimer laser business unit. He holds a PhD in Laser Physics from Braunschweig Technical University and Master degrees in Economics and Business Administration from the Open University in Hagen. the authors Dr. Ralph Delmdahl Coherent GmbH Hans-Böckler-Str. 12 37079 Göttingen, Germany Phone: +49 551 6938 397 Fax: +49 551 6869 1 E-mail: Ralph.Delmdahl@coherent.com Website: www.coherent.com Ulrich Schwarz Prof. Dr. Ulrich T. Schwarz is group leader at the Fraunhofer IAF in Freiburg and Professor for Optoelectronics at the Institute for Microsystems Engineering (IMTEK) of the Albert-Ludwigs-University Freiburg. His research interest is focused on group-iii-nitride based optoelectronic devices, in particular light emitting diodes (LED) and laser diodes. He holds a PhD in physics from Regensburg University. Awarded by the Alexander von Michael Kunzer Dr. Michael Kunzer studied physics at the University of Freiburg, where he received a diploma and a PhD degree. In 1990 he joined the Fraunhofer-Institute for Applied Solid State Physics (IAF). From 1995 on his work is focused on III-Nitrides devices. He has been involved in several LED and laser oriented R&D projects. Since 2009 he is heading the nitride optoelectronic group. Dr. Michael Kunzer Fraunhofer IAF Tullastrasse 72 79108 Freiburg, Germany Phone: +49 761 5159 844 Fax: +49 761 5159 71844 E-mail: Michael.Kunzer@iaf.fraunhofer.de Website: www.iaf.fraunhofer.de Humboldt foundation with a Feodor Lynen scholarship he spent two years (1997 1999) as postdoc at Cornell University, Ithaca, NY. In 2006/2007, he visited Kyoto University, Kyoto, Japan, with an invited fellowship awarded by the Japanese Society for the Promotion of Science (JSPS). Prof. Dr. Ulrich T. Schwarz Fraunhofer IAF Tullastrasse. 72 79108 Freiburg, Germany Phone: +49 761 5159 513 Fax: +49 761 5159 677 E-mail: ulrich.schwarz@iaf.fraunhofer.de Website: www.imtek.de/optoelektronik Figure 1: LED illuminated Yas Hotel in Abu Dhabi. Market Background In 2009, the total market for high brightness LEDs was estimated to be about US$5.3 billion [1]. LEDs for display backlights totaled approximately US$400 million within this, and LEDs for general illumination applications contributed another US$600 million, thus together comprising a little less than 20% of the total. By 2014, the total high brightness LED market is currently projected to quadruple to US$20 billion, with LEDs for backlighting making up half of the total (US$10 billion) and general illumination applications growing to US$4.5 billion.leds 48 LTJ May 2011 No. 3 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

FIGURE 2: Simplified schematic of a traditional LED on sapphire substrate, in which both electrical contacts are placed on the front side. The yellow arrows indicate the current flow. are enjoying this tremendous success (Fig. 1) because they offer several important advantages over other light sources. Some of the most important of these are their compact size, extremely long lifetime, high reliability, physical ruggedness, low voltage requirements and high electrical efficiency. The latter means that they produce little waste heat that helps to maximize battery life in portable devices. In fact, the efficiency (lumen of output/watts of electricity input) of the best high brightness LEDs is already better than nearly every other type of light source, including traditional incandescent bulbs, halogen bulbs, and fluorescent and compact fluorescent sources. Currently, only metal halide lamps, typically used in applications such as street lighting, offer higher electrical conversion efficiencies, but this comes at the cost of an extremely low color rendering index. There are, however, some roadblocks limiting the potential market growth for high brightness LEDs. In particular, lowering cost is essential if the LED is to displace other light sources in both display backlight and general lighting applications. While improving electrical efficiency is also desirable, the best high brightness LEDs are already less than a factor of two away from the maximum theoretically achievable efficiency value. Thus, there s not much more to be gained in this area. To achieve cost reductions, manufacturers are specifically focused on making improvements in LED cost per lumen and in power scaling. That means, increasing the maximum total flux output that can be obtained from a single emitter. The latter factor also relates to cost, since it reduces the number of separate LEDs which must be utilized to produce a given lumen output. But, making substantial improvements in cost/lumen and lumen/lamp may require a fundamental shift in high brightness LED design and construction. Traditional LED Construction FIGURE 3: Simplified schematic of a thin film LED, in which light is emitted primarily from the top of the device. The yellow arrows indicate the current flow. The traditional blue, green or white, high brightness LED consists of multiple epitaxial layers of III-Nitride materials grown on top of a sapphire substrate. The III-Nitride material system, and, in particular, InGaN, is used because it can deliver output in the green through near ultraviolet (UV) parts of the spectrum. Typical cold white LEDs are produced by using a blue LED with a yellow phosphor. Improvements in color quality, such as warm white output, can be achieved through the use of phosphor combinations or UV pump chips. A sapphire wafer of about 300 to 500 µm in thickness is used as substrate material for several reasons. First, as with GaN it has a hexagonal lattice structure. Next, it is transparent at all current LED output wavelengths. This is important because some of the light escapes the LED through the sapphire substrate. Sapphire is also stable, and easily able to handle the high temperatures (up to 1100 C) and harsh conditions experienced during the Metal Organic Vapor Phase Epitaxy (MOVPE) growth of the semiconductor layers. Finally, sapphire wafers are relatively inexpensive. The biggest disadvantage of sapphire is that it is not electrically conductive. As a result, both electrical contacts must be located on the top side of the device. This is depicted in Fig. 2, which shows a simplified schematic of a traditional LED. This configuration forces 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.laser-journal.de LTJ 49

the current to pass laterally through the n- GaN contact layer. The lateral current flow causes an inhomogeneous current distribution through the active multiple quantum well (MQW) region, especially a current crowding at the edge towards the n-contact. Since the efficiency of the active layer is optimized for a certain current density, this effect limits the efficiency as well as the maximum current through the LED device. Since the transparent sapphire substrate is part of the LED design, these traditional LEDs can be considered as 3D emitters. That means, they emit light in virtually all directions, specifically, out of the top surface (depending upon the construction of the contact), through the edges, as well as through the sapphire. This makes the farfield irregular and difficult to implement into an optical design. The traditional LED structure also poses two significant limitations in terms of its scalability. This is important because obtaining higher light output from a single LED usually requires physically increasing its size. The first limitations again relates to the sapphire substrate. Specifically, sapphire is not a great heat conductor. So as device size increases, it becomes progressively more difficult to effectively cool the LED, which is critical to its proper operation and longevity. The second problem is that LEDs on sapphire are limited in scalability due to sizeable contributions of light out-coupling through the edges of the chip. Hence light extraction the company COHERENT Inc. Santa Clara, USA Since its foundation as a laser manufacturer in 1966, COHERENT Inc. has become the technology and market leader in a number of areas. The company, which is headquartered in California, has R&D and manufacturing facilities around the world in Europe these are in Germany and Great Britain. A global service and sales network supports customers in industry, medicine and science. www.coherent.de www.coherent.com European Contact: Petra Wallenta, PR Manager Europe Petra.wallenta@coherent.com FIGURE 4: The steps of thin film LED fabrication include a) Device growth, b) Flipping the chip and eutectic bonding of the bottom to a carrier wafer, c) Exposure to UV laser light, d) Removal of the sapphire substrate, and e) Production of a top side contact. efficiency decreases with increasing chip size. A third practical disadvantage of sapphirebased LED wafers is the elaborate dicing process required to singulate them. Traditional semiconductor substrates, such as silicon, can be cleaved. Due to its very hard and brittle nature, sapphire must first be thinned to less than 100 µm and then sawed. In both cases, slow and expensive diamond-based processes have to be applied, making the singulation process an important cost factor in LED processing. Thin Film LEDs Increasing device size is necessary to obtain higher light output from LEDs, but, clearly, it is difficult to scale up traditional LED structures because of the architecture limitations imposed by the use of sapphire substrates. To solve this problem, a new design and manufacturing approach, called thin film construction, has been developed for larger size, high brightness LEDs. Thin film LEDs consist of a number of epitaxial layers grown on sapphire, just as is the case for traditional LEDs. However, in this instance, the structure is configured as a simple stack of layers. The current flows vertically from the top to the bottom of the device (Fig. 3). After epitaxial growth, the chip is flipped, and a carrier wafer of choice is bonded to the now bottom (p) contact of the LEDs. This carrier wafer consists of a material which provides both good electrical contact and heat dissipation characteristics. Typical are silicon, germanium or a metal. LLO is then employed to remove the sapphire substrate. Next, n contacts are deposited on what is now the top of each LED die. Finally, the individual LEDs can be easily singulated (Fig. 4). The thin film LED structure solves most of the problems encountered in traditional sapphire based LED designs. First and foremost, it can be scaled up in size without difficulty. One reason for this is the mainly vertical current flow, which eliminates most current crowding issues. Furthermore, having a substrate with high thermal conductivity enables heat to be effectively dissipated, even from a large device. So it s quite possible to construct single, thin film LEDs having a chip size of several millimeters, which still deliver high electrical efficiency and excellent device lifetime characteristics. Another significant advantage of the thin-film LED configuration is that most of the light exits the front surface of the device. This is because the edge region becomes a negligible fraction of the area of the LED with increasing size. For example, if the emitting region is 5 µm thick, and the LED is 2 mm x 2 mm, then the front surface is much larger than the combined area of the edges. The light output of the latter can be essentially ignored. Also, because most of the emitted photons are traveling a relatively short distance before exiting the device (rather than long distances laterally through the material), reabsorption ceases to be a problem. To further enhance the forward emission characteristics of thin film LEDs, the p-side contact is usually formed from a highly reflective material, such as silver. Also, the epitaxial layer stack can be designed to be a resonant cavity which enhances emission towards the front. As a result, with most of the light now emitted from the front of the device, it becomes far simpler to efficiently collect and use this light using even simple optics. One general problem with LEDs is trapping of light inside the chip due to total reflection. Since there is a very large index of refraction drop from the epitaxial layers to the surrounding environment, the angle at which total internal reflection (TIR) starts is fairly shallow. This problem can be solved 50 LTJ May 2011 No. 3 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

FIGURE 5: The COMPexPro based micro- STRUCT LED system is a complete tool for LLO process development and pilot production. by microstructuring or roughening of the emitting surface of the LED. Thereby exiting photons mostly encounter the surfaces at small angles of incidence, and therefore exit with minimal internal reflection, and mostly avoid experiencing TIR. Such a microstructuring can only be achieved after performing substrate removal, and is a key to achieving high efficiency in thin-film LEDs. FIGURE 6: Standard deviation (Sigma) of pulse energy vs. total pulse count measured over three billion pulses. Each point represents data from 1000 pulses. Laser Lift-Off Separating the epitaxial layers from the sapphire substrate is one of the critical steps in thin film LED production. LLO is the method of choice for accomplishing this separation. In LLO, the LED wafer is exposed to high intensity UV light from the sapphire substrate side (Fig. 4c). Sapphire is transparent in the UV, so this light simply passes through it. However, GaN is highly absorptive in the UV, so the light is absorbed in an extremely thin interface layer (~50 nm in thickness). If a short pulse laser is used, there is no risk of optical or thermal damage to the underlying LED structure from this surface effect. And, if the incident energy density is sufficient, the material is decomposed into metallic gallium and nitrogen gas. Gallium is liquid 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.laser-journal.de LTJ 51

above room temperature, so it can stay in this state indefinitely. The LLO process works by sequentially exposing areas of the wafer to high energy pulses of UV laser light until the entire surface area is covered. Once the whole interface layer has been liquefied, the sapphire substrate can be slid off, leaving the LEDs only bonded to the carrier wafer. But there are critical requirements that a laser source for LLO must meet. First, it must be able to emit light with a wavelength of 350 nm or less. Furthermore, it has to deliver an energy density at the sapphire/gan interface of 600 mj/cm 2. A value lower than this is not sufficient to decompose the GaN. On the other hand, too high energy densities produce a shock wave in the nitrogen gas that can crack the devices. In practice, an area covering several LEDs is illuminated at once, and the energy density must be maintained within a few percent over this entire region, so beam uniformity is critical. Due to these requirements, the excimer laser, especially the KrF excimer laser operating at 248 nm, is the first choice for LLO. This is because the excimer laser delivers a high energy pulse that is naturally rectangular in shape with uniform intensity distribution. This beam can readily be manipulated with beam shaping optics to match the necessary processing area on the work piece, and to have a homogeneous intensity distribution over most of its area. These requirements would be much more difficult to achieve with other UV laser technologies, such as the diode-pumped solid-state laser. They typically deliver much lower pulse energies and have a Gaussian beam profile that is harder to transform into a uniform and rectangular flat-top profile. the institute Fraunhofer Institute for Applied Solid State Physics The Fraunhofer Institute for Applied Solid State Physics (IAF) is a leading research and technology center in the field of micro- and nanostructured compound semiconductors and Diamond. Its research work is focused on the development of micro- and optoelectronical circuits, modules, and systems. http://www.iaf.fraunhofer.de The Fraunhofer Institute for Applied Solid State Physics together with 3D-Micromac, has designed a tool for LLO, called the micro- STRUCT LED system (Fig. 5). This system incorporates a Coherent COMPexPro 102 KrF excimer laser (400 mj at 20 Hz output) and uses Coherent beam shaping and delivery optics to transform the 24 mm x 10 mm output beam from the laser to a highly homogeneous pattern with the necessary energy density for LLO. It also integrates a granite base for stability, together with a wafer motion platform, to enable LLO of LEDs on sapphire substrates of up to eight inches (200 mm) in diameter. The 20 Hz COMPex- Pro laser-based system is primarily optimized for process development and pilot production applications. In use, the system gets first adjusted so that the beam dimensions are an integral multiple of the LED die size. This allows the edge of the laser illuminated area to always be placed between two LEDs. Then the wafer is exposed to a series of pulses, and is moved a distance corresponding to the width of the illuminated area between each exposure. This exposure pattern prevents minor mismatches that might occur at the overlap area between successive exposures from adversely affecting the active area of the devices. The system can process a two inch wafer using a field size of 2 mm x 2 mm in one minute using this approach. Laser and Process Upscaling Considerations Key requirements for the excimer laser in LLO are pulse-to-pulse stability and a consistent beam intensity distribution. In order to guarantee temporal and spatial beam homogeneity over the entire life of a laser discharge unit, several design advances have been implemented in high power excimer laser models. These are proprietary solid state switching devices, optimized high speed gas flow management and highly advanced electrode materials and shapes. The enhancements result in an extremely high pulse-to-pulse stability maintained over several billion pulses of high power laser operation, as shown by the standard deviation (Sigma) in Figure 6. The energy stability in the graph was achieved with a 130W LEAP model at 248 nm and 650 mj pulse energy over three billion laser shots, and is far better than the 1% required to stay within the LLO process window. Furthermore, virtually all maintenance has been eliminated in advanced high power excimer lasers, except for automated gas changing, which takes only a few minutes. Thus, maintenance costs and downtime for these excimer lasers rival with solid-state lasers. Production floor LLO processing would ideally like to achieve sufficient speed to process a six inch diameter wafer in one minute. This is about a factor of ten faster than the above mentioned system. Excimer laser technology at 248 nm is available over a large range of pulse energies and repetition rates. It is therefore an enabling technology, allowing scale-up of throughput via both illumination field size increases and process speed increases [2]. In conclusion, high brightness LEDs are a vital component in an expanding array of consumer electronics devices and lighting products. However, the cost and performance drawbacks of traditional LED designs could limit the expansion potential in some of these markets. Thin film LEDs, produced by using 248 nm excimer laser based high throughput LLO processing systems, promise to overcome these limitations and enable these markets to reach their full potential. References [1] R. Haitz, J. Y. Tsao: Solid-state lighting: The case 10 years after and future prospects, Phys. Status Solidi A 208, 17 (2011) [2] R. Delmdahl, R. Paetzel: The Midas Touch: Surface processing with the UV excimer laser, Laser Technik Journal 1/2009, 24 52 LTJ May 2011 No. 3 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim