LED Lights: Emerging Opportunities
Dr. Wessner introduced and congratulated Mr. Trimble for his major contribution to the development of the global positioning system and for his service on the steering committee of the task force on Government-Industry Partnerships.
Mr. Trimble said that he felt like Rip van Winkle in moderating this panel. The last time he had studied LEDs was in the late 1970s, when he worked at Hewlett-Packard. At that time, his company was unable to convince the auto industry of the value of using LEDs for brake lights. Today, he said, the efficiency of LEDs has increased by two orders of magnitude and the auto industry is embracing them rapidly.
THE EVOLUTION OF LED LIGHTS
The Heart of the LED
Dr. Craford began by describing the basics of LED technology. The heart of the system, he said, is the LED chip itself, which is a semiconductor com-
posed of different layers. The material used in each layer determines the color of the light, and growing the layers carefully affects the amount of light extracted. To make red, orange, and amber LEDs, LumiLeds uses aluminum gallium indium phosphide, or phosphide for short. For blue and green, LumiLeds uses a newer style of chip made of gallium indium nitride, which is grown on a sapphire substrate.
Two central issues, he said, are light generation within the chip and light extraction from the chip. If the first stage is done correctly, light can be generated inside the chip with nearly 100-percent efficiency (i.e., every electron generates a photon). Light can also be extracted with an efficiency of 100 percent, in theory. In practice today the highest light extraction efficiency is about 50 percent. There are many practical barriers to higher extraction efficiencies, including internal reflection and absorption at the metallic contacts, within the semiconductor layers, or in the encapsulation/package.
A History of LEDs
General Electric sold the first LEDs in 1962, based on the work of Nick Holonyak Jr., then with General Electric Research laboratories. They were comparatively dim and very expensive, costing several hundred dollars each. Since the late 1960s, when Monsanto and Hewlett-Packard started marketing LEDs in high volume, their light output has increased by about 10 times per decade. The brightest LED today, he said, produces about 100 lumens per watt (lm/W). He demonstrated one of the small, dim, red diodes made 30 years ago and compared it with a much brighter 100 lm/W LED flashlight. “We’ve come a long way in red,” he said.
The value of LEDs can be seen easily in some of their red applications, such as brake lights and taillights. The early Edison bulbs emitted about 2 lm/W; modern incandescent lamps emit around 15 lm/W. When a red brake light is created by putting a red filter over an incandescent bulb, most of the light value is lost, and the output drops to 3 to 4 lm/W. With a red LED, virtually all the light output is functional.
As a result, most of the car manufacturers in the world now use LEDs for their high, center-mounted brake lights. In the Cadillac Deville all the tail lights and brake lights are LEDs, which are not only brighter but can also be mounted in smaller spaces in the car. As a result no space is lost to house bulbs and fixtures. They cannot be seen at all in some applications unless they are turned on. LEDs have also begun to replace neon for signage because of their energy efficiency, reliability, and lack of toxic mercury.
The other successful application discussed at the symposium was traffic signals. The early solid-state models held about 700 LEDs. As brightness improved,
this number fell to 200; today’s signals contain only 10 to 20 high-powered LEDs. Each signal head uses 15 watts and has a lifetime of 120 months; by comparison, incandescent bulbs use 135 watts and last about six months, a 9x energy savings and a 20x maintenance savings.
While the natural advantages of LEDs are easily suited to these early applications, the next stages of application will require major technological advances. Dr. Craford summarized some technological trends in LEDs that are leading to the next generation of applications.
Chips are getting larger. In the early 1990s, the standard phosphide chip was 0.25 mm square, drawing just 0.1 watt and putting out a few lumens of light. Now a bigger chip, 0.5 mm square, emits tens of lumens. In the newer gallium nitride technology the trend is similar, from smaller to larger, from low to high power; GaN chips now emit up to 50 lumens, and in some research devices, over 100 lumens.
Chips are being shaped. LED chips, which have a high index of refraction, suffer from the problem of light trapping, which limits the extraction of light from the chip. Recently, chips with tapered sides or roughened surfaces have become available. These shapes break the symmetry of the chip and allow more light to escape.
More white-light LEDs are being made. White LEDs have generally been 5 mm in diameter and put out about a lumen of light. These LEDs are not made by mixing red, green, and blue (R-G-B) layers, but by starting with a blue chip and adding a yellow phosphor.1 This is a simple system, but some efficiency is lost by going through the phosphor. The superposition of R-G-B LEDs to produce white light can yield higher performance, but the blue light/phosphor system is simpler and more convenient for low-power uses, such as in glove compartments and stairway night lights. Higher power LEDs emitting around 20 lumens were recently (late 2001) introduced commercially.
The Major Challenge of Cost
High cost is a major challenge for commercialization. The selling price of a 5mm white-light LED is about 50 cents to $1 per lumen; that same dollar can buy four 100-watt tungsten bulbs. Each bulb emits 1,500 lumens. We thus have up to a 6,000-fold difference in lumens per dollar. Considerable work must be done in manufacturing technology and other fields to drive that cost lower. Costs have been dropping about 10× per decade, and light output per package has been rising by 30× per decade. The rate of cost reduction needs to be accelerated to enable
LEDs to compete effectively with conventional light sources for high power applications.
Other Barriers to Commercialization
Several other issues continue to stall commercialization:
Reliability. Red and amber LEDs are very reliable, but the long-term reliability of short-wavelength blue and white has not yet been demonstrated. The devices must be stable at high current density drive conditions to enable the transition to cost effective high power applications.
Packaging. The package surrounding LEDs has to let light emerge more efficiently, be stable against blue and UV radiation, and in high temperature, high humidity environments.
Energy savings. The level of reliability, efficiency, and cost of some LEDs make related energy savings more prospective than real. White LEDs will probably always cost more than traditional bulbs, so they must earn back these costs through low energy consumption. In their favor is that incandescent and fluorescent bulb efficiencies are not improving much with time, while LEDs are improving steadily.
Power conversion efficiency. In combination with red, green, and blue, a 45 percent external quantum efficient LED is not far from 50 percent, because a white light of 150 lm/W could be realized if 50 percent was achieved at all three colors. A red LED with external quantum efficiency of 45 percent, depending on peak wavelength and bandwidth, would have an efficacy of less than 100 lm/W. For the other colors, efficiency falls off. For the nitride system, efficiencies are around 25 percent for UV, 20 percent in the blue, and 10 percent or lower in the green range, so that efficiencies have to improve by a factor of 3 to 5. There is no known theoretical barrier to this goal, but no one knows how long it will take.
In summary, said Dr. Craford, LEDs have advantages in monochromatic applications (such as brake lights). Their use is increasing in those applications, and eventually they will become the standard. They will also be increasingly used for low-power applications, especially where reliability is paramount or bulb changing is difficult, as in tunnel lighting. Before LEDs can penetrate the general illumination market, they need major performance improvements, cost reductions, and better reliability. “Fundamentally, all this is possible,” he concluded, “and given enough time, it will happen.”
NITRIDE LIGHT SOURCE: BLUE LEDs
Cree Inc., and University of California at Santa Barbara
Dr. DenBaars identified nitride LEDs as “the light engine for advanced lighting.” He is the scientific advisor to the company Cree, Inc., whose main business is to manufacture high-efficiency, gallium nitride, solid-state light emitters.
Completing the LED Spectrum
The opportunity to make white light was not available until the breakthrough in blue LEDs, he said, which emerged from the pioneering work of several professors in Japan. The ability to produce blue completed the color spectrum, after scientists at Hewlett-Packard had introduced both red and yellow LEDs. Gallium nitrate technology, which led to the development of blue LEDs, now makes it possible to create both white light and high brightness green light. The more recent development of ultraviolet emitters adds a new way to produce white and very deep ultraviolet sources. He called the ultraviolet emitters a unique material with many uses in defense (e.g., for ultraviolet chemical and biological agent detectors and for microwave power transistors). He said that the U.S. government is funding some of this work, but that solid-state lighting itself is receiving as little as a few million dollars per year in public money. He noted that in Japan, Taiwan, and Korea—the other countries leading this effort—research receives substantial funding from their governments.
Moving Toward White Light
“How do we get to white lighting?” Dr. DenBaars asked. He said that the most urgent need is to raise the “wall-plug efficiencies” as well as the luminous output. He added that the flashlights shown at the symposium were “the brightest [LED] sources I’ve ever seen, and if we can scale that up another factor of ten we can get to the white-light bulb.” Also needed are substantial improvements in lifetimes and color rendering.
One of the most challenging limitations of solid-state lighting is the cost per lumen, which is very high compared to other sources. However, government support has clearly improved the technology and has lowered the costs of silicon and gallium arsenide technology; a similar federal effort could accomplish the same objectives for solid-state lighting. Other limitations are low color stability, scaling problems, and packaging needs.
A Demand for Greater Efficacy
He showed a plot of efficacy in terms of lumens per watt. He said that some green-light devices had been able to achieve 90 lm/W, and he suggested that
“with modest increases we can be in 150 lm/W range.” This would move solid-state lighting into many demanding applications, including streetlights, so that attaining 150 lm/W would be “a game-changing opportunity.”
Benefits of Solid State
The benefits of shifting toward solid-state lighting are “enormous,” he said, “if we can get to 150 lm/W efficiency and replace 55 percent of existing light sources.” He noted Dr. Haitz’s calculation that full implementation of solid-state lighting would virtually replace 93 nuclear power plants, and he cited other estimates of as many as 133 power plants, which would have a “huge impact on the environment.” This level of energy replacement could reduce global energy consumption by 10 percent, reduce carbon dioxide emissions by 200 million tons per year, and save $100 billion in electricity costs and 1,000 terawatts of power. It would also bring the United States into compliance with the Kyoto environmental accord.
The Complexities of Nitride Technology
Thinking about the place of solid-state lighting in the future, he predicted that nitride sources would be used to generate ultraviolet and blue light, but enormous work was still required to understand the gallium nitride (GaN) nanostructures. For example, after seven years of research it was still not known why it was possible to create efficient GaN LEDs with dislocation densities exceeding 108, other than it probably “has something to do with the fundamental nature of GaN nanostructures.” All other LED materials require very low dislocation materials. Nor is there yet a satisfactory GaN substrate. Current practice is to put the LED on very lattice-mismatched systems using sapphire or silicon carbide. This difficulty is preventing wide implementation of GaN LEDs at very high currents.
In addition, much work is required in the use of phosphor materials to create white light. Investigation in this area is being funded by NIST through an Advanced Technology program with General Electric, Cree, and GELcore, for the purpose of finding alternative ways to generate white light. For the time being, the commercially available nitride process (which uses a blue LED plus a yellow, garnet-based phosphor) gives a white light, but its color rendering is poor. One experimental approach would be to use ultraviolet or short-wavelength blue LEDs with three types of phosphors (pigments, dyes, and polymers) to yield a white light with good color rendering. Dr. DenBaars predicted that both of these technologies, perhaps with one of them using three red-green-blue LED chips to get white light, will all find different applications in niche markets.
A principal issue to address in nitride technology is the same as it is in silicon and gallium arsenide (GaAs) technology. Process technology is immature; all manufacturers are growing their LEDs on 2-inch wafers, which are too small to be economical. Recently the GaAs process migrated to 6-inch wafers, bringing
significant cost reductions; silicon carbide is now grown on 12-inch wafers. Government funding supported the wafer programs, and more of the same, he said, is needed for GaN. If the progress with GaAs is a reliable indication, it will take 5 to 10 years to develop a 6-inch wafer for GaN.
The Filament of the Future
Cree has announced a performance achievement of up to 32-percent quantum efficiency with GaN. In actual watts emitted, these devices gave a wall-plug efficiency of 27 percent. He predicted that the ultraviolet and blue chips would eventually be the building blocks for the high-lumen-per-watt LED lamps, or the filament of the future.
Reaching higher efficiencies requires novel chip concepts. Cree has developed ultra-bright chips using a new light-extraction technique. In contrast to conventional square chips, where light “rattles around inside,” Cree has been able to enhance efficiency significantly through chip shaping. This design redirects more of the photons out of the chip. The lesson here is that simple changes in manufacturing processes can lead to large benefits in brightness.
The Problem of Overheating
When more powerful LED lamps, such as larger flashlights and room lights, are built, the problem of overheating arises. Engineers have found they can reduce some of the heat by pulsing the light instead of maintaining a constant light. They are now able to reach an efficiency for large chips of about 1.2 watts of ultraviolet light for 9 watts of power flowing in; this corresponds to a 15-percent wall-plug efficiency. This is approaching the range of a small incandescent bulb; a 25-watt incandescent bulb, for example, puts out about 1.3 watts of visible light. Nevertheless, thermal management will continue to be a challenge.
Developing New Markets
Dr. DenBaars predicted that additional markets for LEDs will develop quickly as the technology improves. The flashlight market is promising, because LEDs can reduce the relatively high current cost of batteries. The cost point for an LED flashlight might be around one dollar. The major target market of halogen and incandescent bulbs, however, can be penetrated only after substantial technical breakthroughs. The major challenge is to raise the luminous output by at least an order of magnitude. He visualized two possible commercial modes. One is to fashion drop-in or screw-in replacements for bulbs.
The second challenge is to find more ingenious and appropriate ways of surface mounting or embedding the chips at the locus of application, such as in architectural tiles or dome lights. Embedding saves huge amounts of space in such crowded locations as automobile dashboards, as well as reducing heat output and failure rate. Another potentially enormous market is the backlights for computer notebook displays, where LEDs promise great power savings and extended battery life.
Two Major Challenges: Cost and Performance
Dr. DenBaars summarized the primary technical challenges under two headings: cost and performance. He said that nitride LEDs would be the primary light engine for solid-state lighting technology, once its performance is raised from the current 20 lm/W to 150 lm/W, an improvement “I think we can achieve in a five-year time frame.” Also needed is a total output flux of about a thousand lumens for room lighting, which will be driven by large-area epitaxy development and large-area substrates.
It is too early to know whether bulk GaN will eventually be developed, he said; if not, the technology will have to be developed on silicon carbide and sapphire substrates. It is not clear which of these two substrates will win out. Although the technology and its merits have been proven, engineers now have to scale it up. He predicted that once the industry’s primary performance goal is attained, “everything will be there. Once we get to 150 lm/W, it will pay for itself in a one-year time frame. That’s the message I want to bring. But, we need a big breakthrough to get there, and a focused effort.”
AVENUES TO WHITE LIGHT
National Institute of Standards and Technology
Dr. Gebbie began by describing a conference convened by the Department of Energy in 1998 to discuss the future of lighting and develop a roadmap for the next 20 years. “It was not surprising,” she reported, “that in considering solid-state lighting, industry focused part of its attention on the need for objectives, definitions, and standards for lighting quality.” Indeed, she said, it has often happened during major technological changes that traditional methods for measuring quality are found to be outmoded and must be replaced by new ones. Producers and customers often find large discrepancies in their measurements without understanding why, and this undermines the confidence of participants in the marketplace.
Defining and Measuring Lumens
Dr. Gebbie said that NIST can help reduce such problems by working with manufacturers to address their measurement needs and to coordinate standards internationally. In describing a series of efforts under way, she noted that while the lumen is based on a well defined measurement of the photopic response of the eye, the perception of color and the quality metrics of color temperature and color rendition are based on measurements using broadband emitters. These metrics are being questioned as combinations of narrow band emitters, qua LEDs, are being created to provide white light.
Difficulties in Measuring Light
Standards that are based on human senses present a problem: A human is needed to be part of the measurement process, but one human may make a different judgment from another. Even the same human will make different judgments at different times. One of the greatest achievements of NIST in the first part of this century, she said, was to create a standard model for human vision. This model, worked out by Gibson and Tyndall in 1923 and fortified by data from some 200 people, determines the relationship between luminous intensity and the visual response of the eye. The model defined the response as a weighted average over a number of wavelengths or colors. Weighting is necessary because, for example, green contributes the most lumens, blue and red fewer.
This functionality model represented an enormous advance in the measurement of luminosity. In 1924 the International Commission on Illumination adopted it as the world standard, and in 1933 the International Committee on Weights and Measures, which is responsible for the metric system, followed suit. In 1948 world organizations for the first time accepted a single standard of brightness (based both on the emission of red-hot platinum at its melting point) and on the response of the human eye. Finally, in 1979 the International Conference of Weights and Measures adopted the present system, which is based on the watts emitted by a lamp. This system still uses the original visibility curve of Gibson and Tyndall.
The 1979 system made it possible to design and build an entirely physical, electrical instrument to measure the brightness of light; NIST’s Photometry Standards Laboratory has developed such an instrument. However, this model, she said, still has limitations. For one thing, the world of lighting has changed markedly since 1923, when there were no narrow-band light sources, such as phosphors and LEDs. The human eye may respond very differently to narrow-band lighting than it does to broadband light sources. Nor does the model always predict the accurate brightness even when measurements are made correctly.
Taking Human Factors into Account
Two points, she said, should be emphasized. One is that small differences of
10 to 20 percent in lm/W have no significance in the general appearance of light to humans. Second, if we are to have an understanding of the effectiveness of lighting, human factors must be taken into account and vision experts must be included in research. Two different people measuring the same light can give responses that differ by 50 percent, which is much greater than can be explained by physical differences. Even without including human responses the new demands of spectral distribution require new instruments and methods.
NIST is working to resolve these problems. She described plans to host the second CIE International Commission on Illumination (CIE) symposium at NIST (the first was held in Vienna in 1997 before the recent breakthroughs for green and blue LEDs). In addition, the CIE has established technical committees to develop uniform standards; NIST participates in this process, representing the interests of U.S. industries.
“Good” Light Versus Low-CRI Light
Major ongoing issues revolve around understanding and measuring the composition of light. Isaac Newton demonstrated that white light is composed of a continuous rainbow of different colors; this can be demonstrated by the “rainbow” output of a modern diffraction grating. The reverse is not necessarily true, that is, not all white light is made up of a continuum of all colors. For example, a white light on a TV or computer screen is produced by superimposed pixels of R-G-B light. To incorporate and describe these differences industry has developed the Color Rendering Index (CRI). The highest color rendering value is 100, which may be seen in full sunlight. Values over 80 are considered good. The CRI index for fluorescent range from 62, which is acceptable, to 95, which is high. In general, broadband lighting has a higher CRI and looks more natural than narrow-band lighting. NIST’s photometry lab is helping to explore ways of raising the CRI of LEDs.
The Need for Better Measurement
Dr. Gebbie summarized her talk by reminding participants of the urgent need to develop uniform ways of measuring and describing the performance of LEDs. The marketplace must understand the new forms of light in terms of quality, comparability, and consistency as well as performance. This understanding, she concluded, would have to involve not only better physical measuring tools but also better knowledge about the responses of the human eye to various combinations of wavelengths.
CRITICAL R&D CHALLENGES
Dr. Karlicek said he would review the discussion of R&D challenges that had been raised at the symposium. He began by saying that government support would help address these challenges, especially in view of the competing countries in Asia. Some of these countries have large government-funded initiatives to advance solid-state lighting.
Challenges to the Lighting Industry
He summarized three challenges that faced the industry in its effort to develop white LEDs:
The first is performance, or lumens per watt (lm/W). Traditional lighting sources emit 10 to 100 lm/W; LEDs in the research lab have been reaching about 25 lm/W. A single traditional white-light bulb emitted a kilolumen or more. This, he said, presents a “rather large performance challenge.”
Costs need to come down. Traditional bulbs cost about a dollar per kilolumen, LEDs around $500 per kilolumen.
Packaging, he said, will be a “gating factor,” not only in terms of efficiency but also as an interface between application and fixture. A lot of work has not been done on packaging.
The technical components of white light can be described under three categories: the LED chip, LED packaging, and the phosphor.
Improving Chip Technology
The big LED cost driver today is the LED chip. One major challenge is that the substrates are very small, and many of them are needed to make large quantities of chips. Substrate size must be increased to produce more chips per wafer. This will become even more important when using larger chips to generate more lumens.
The epitaxy process by which the layers of the chip are grown is a very complex chemical process characterized by large and expensive capital equipment, fairly low throughput (a few wafers per day), and low yield.
Better Chip Fabrication
LED chip fabrication, although similar to the process used to grow chips on silicon, is a manual process. It is not amenable to automation because few compa-
nies make equipment that can handle 2-inch substrates. The industry needs larger substrates that are lattice matching. Fabrication of GaN substrates is a technical challenge. Most companies are attempting to use heteroepitaxy, with sapphire or silicon carbide substrate, but some people hope to use silicon wafers with an innovative lift-off technology that would allow the substrate to be peeled off the epitaxial film and either reused or discarded. This would produce large areas of useable chips without the substrate.
Problems with Epitaxy
In the epitaxy area reactor design will be critical. The industry needs to be able to grow high-quality materials in huge volumes (square meters instead of square inches) and it needs better understanding of chemical and flow modeling inside the reactor to improve the yield. One 2-inch sapphire wafer can produce about 14,000 small chips, and yet this may represent a yield of only 20 to 30 percent; most of the product is being thrown away. Substrate and epitaxial growth are two areas where major breakthroughs will be required to move cost performance to a commercially viable level.
Chip performance also must be improved. All photons originate inside the chip, and the challenge is to direct them outside the chip with high efficiency. At present, he said, his lab is using a gallium nitride crystal structure characterized by fairly high defect densities. There are also doping problems that do not allow optimal electrical conductivity in these materials. This is one reason for the yield problems on these wafers. Extensive research is needed on better lattice matching, heteroepitaxy, and defect modeling.
Another issue in chip performance is efficiency. A gallium nitride chip on a sapphire substrate is very efficient at converting electrons to photons at low voltages, but it is not very bright. At high currents the efficiency is very low. Improving chip efficiency depends on defect reduction, novel designs, and large, efficient chips.
The challenge of light extraction (guiding the photons out of the chip) involves many of the same issues. Better light extraction requires innovative fabrication techniques, such as chip shaping. This possibility has been understood since the first LEDs were made in the 1960s, but more effective shaping is needed. This may also require either a lattice-matching substrate or improved understanding of how to do heteroepitaxy, where the GaN is grown on different kinds of materials to drive down the cost. Also needed is better modeling of defects and doping performance for these materials to understand their basic atomic structure.
To summarize, chip costs will have to be reduced by 10 to 50 times to be competitive in the marketplace. Major advances are needed in substrate technology, in heteroepitaxy on large wafers, in epitaxial yield, and in wafer uniformity. At the same time, chip efficiency must be increased by three to five times through improved structure design, higher efficiency in materials systems regardless of wavelength, and novel fabrication devices and new structures for better light extraction.
Problems of LED Packaging
An area that is frequently overlooked in high-efficiency LEDs is the package. Since the development of the 5-mm LED lamps of the 1960s, packaging has steadily improved in its ability to tolerate heating and to conduct heat out of the package to the environment. In spite of the improvements thermal conductivity must be raised by a factor of two or three beyond what is now possible. Most LEDs use an epoxy system of encapsulation that does not respond well to high temperatures or to blue or ultraviolet light. Most epoxies absorb a certain fraction of that light and degrade very quickly, letting in destructive moisture and air. This degradation is one reason why white and blue LEDs have short lifetimes.
The challenges to improving packaging are minimizing the heat produced per watt, conducting more heat out of the package, and designing packages that are thermally stable. Thermal management issues are critical for the lifetime, lumen output, and fixture design of high-lumen LEDs. Fortunately, some of the research needed on packaging will also support other areas of semiconductors, including power electronics and applications for communications and military systems.
To improve packaging and increase chip lifetimes, the following are needed:
A higher index of refraction to match the chip index of refraction, so the light can be extracted from the chip and into the environment;
Greater transparency of encapsulants to ultraviolet light to avoid photodegradation by the light generated by the chip;
A good match between the chip’s coefficient of expansion and that of the package and the materials that hold the chip in place;
Good adhesion and low moisture permeability; and
Development of new polymers or copolymer systems as encapsulants.
Another LED performance driver is phosphors. For chip-plus-phosphor technology, light is generated by the semiconductor and then converted by the phosphor to visible light. Thus the phosphor has two goals: (1) to absorb the light
efficiently and (2) to generate the light at a different wavelength with high efficiency. This calls for phosphor systems that are excitable in the ultraviolet or blue with high efficiency. The phosphors must also be packaged in materials with good thermal and chemical stability. The light must cover the entire visible spectrum with good efficiency to create the high-CRI illumination systems that will be required for high-quality white light.
When three (R-G-B) chips are used to generate white light, instead of a chipphosphor combination, an additional complexity arises. For some materials the correlation of chip output and temperature varies according to color. For example, the efficiency of the blue output goes down with increasing temperature while the efficiency of red-yellow rises. Because of such variations in temperature performance, the system will require not only better thermal performance but also some color-correction circuitry to be able to tune the color. While such circuitry adds expense, it also allows the user to control the color as desired.
Sector Strengths: Academia, Industry, Government
In describing the overall R&D effort needed to produce white LED light, Dr. Karlicek offered a breakdown of the contributions of the academic, industrial, and government sectors. For improved chips, for which device design and manufacturing breakthroughs are needed, the breakdown in funding is roughly equivalent for all three sectors. This is because a high amount of industrial funding is required to develop certain technologies that are close to commercialization, and at the same time multiple breakthrough technologies are needed to bring the needed performance. For improved packaging design, which affects both performance and cost, both new optical polymers and breakthroughs in thermal performance are needed. This will require large contributions from government labs and industry and a smaller contribution from academia. For improved phosphors, which affect chip performance, the need is for new high-efficiency phosphors. Here the major effort must come from industry, with a smaller contribution from academia and minor input from government labs.
The Importance of Design
Finally, he compared LED lighting with traditional lighting and emphasized the importance of design. By properly designing a lighting system a relatively low-lumen LED can actually generate more useful light than an incandescent bulb. He showed a wall-wash application that traditionally might be lit by an 11watt incandescent with 150 source lumens. Illumination that is roughly two to three times more efficient can be achieved by an LED system using only 1 watt of power and 10 source lumens. A second advantage is that the LED is cool while incandescent burns hot.
Box A. Hierarchy of Needs
Dr. Karlicek summarized his talk by ranking the following needs in order of urgency:
In all, commercialization of LED lighting depends on these system challenges:
Increasing the total lumens per lamp;
Developing uniform sockets;
Developing application designs for LED requirements, where the LED and the application are designed to fit one another; and
Adapting LEDs, which basically run on direct current, to a power supply and distribution system designed for alternating current.
“The bottom line,” said Dr. Karlicek, “is that we need major breakthroughs in lighting paradigms to be able to use solid-state lighting and achieve the energy saving benefits they offer.”
Charles Trimble observed that if one looks at successful partnerships between academia, government, and industry in the computer, semiconductor, and Global Positioning System arenas, one finds several common threads. First, the government invested seed capital in precompetitive research. Second, the government provided early markets for the products. He suggested that the panel address the subject of white-light/LED generation and identify the three most important areas of precompetitive research that are needed. The panel offered the following responses.
A Need for Breakthroughs
George Craford of LumiLeds Lighting said that the nitride technology had come up to a certain level fairly rapidly, but progress had slowed over the last few years. He said that this plateauing is a familiar phenomenon in the development of new technologies. From this plateau the industry still has to improve chip efficiency from three to five times, depending on the color. He said that he had not considered where to draw the line in defining precompetitive technologies, but genuine breakthroughs in efficiency “certainly must happen to allow this lighting technology to take off.”
Steven DenBaars stressed the need for better understanding of the physics of light emission, including the characteristics of nanostructures, the role of defects, and the high voltages seen in the nitrides related to p-doping. “The basic materials science is not well elucidated,” he said.
Katharine Gebbie emphasized the need for extensive research in measurements, both precompetitive and infrastructural. Better metrics will benefit all of industry, as well as customers, and will be “absolutely essential” for the field of LEDs to complement technological advances.
Better Knowledge of Chemical Systems
Bob Karlicek said he agreed that the physics of defect analysis of semiconductors and gallium nitride would improve understanding in a precompetitive sense. He also urged a focus on some of the fundamental growth and chemistry issues that go into heteroepitaxy, given the poor likelihood of attaining 6-inch gallium nitride substrates “in my lifetime.” Defect analysis, he said, will be a challenge because it requires learning how to drive costs by doing heteroepitaxy more effectively. He also said that infrastructure support, although it is not really precompetitive, would be critical. The bottom line is that more needs to be known about the very complex chemical systems that drive the costs of the LEDs. Having better growth technologies and manufacturing technologies for both fabrication of chips and especially for the epitaxial growth process would help us reduce costs to a significant extent. Unlike the silicon industry, which is a very large business that can pay for a great deal of R&D on equipment, the three- to five-compound semiconductor business is not large enough to do that yet, and the chemical and physical challenges to the growth and fabrication processes are much more complex. The funding challenges are exacerbated by a combination
of two factors: an increase in complexity of compound semiconductors and the current small size of the market.
Is Government Support Needed?
Pat Windham asked a follow-up question about funding needs. He cited a perception in Washington that the new, high-tech industries involved in such fields as solid-state lighting are already doing enough technology development, thanks to partnerships with the venture capital industry and large companies. According to this view there is little need for government funding beyond some support for basic research at universities. What, he asked, is the rationale for a larger government role?
George Craford responded that part of the reason for government support is a matter of timing. If the industry is allowed to evolve without any outside stimulus or national policy, he said, it would evolve slowly at a time when the governments of other countries are aggressively funding research. Private firms are limited in how much they can invest in research by the realities of quarterly and even monthly statements, and their R&D money is usually applied to the shortest-range research on new products to generate revenue. The primary beneficiary of high efficiency LED lighting will be the consumer who will save on energy costs. The LED companies will also gain revenue but general LED lighting is perceived to be quite far in the future and is high risk. Given enough time, he said, the industry will eventually develop the technologies needed, but if the nation wants to accelerate that development, it will require some government push in basic research and in development.
Dr. DenBaars pointed out that the industry would not have 6-inch gallium arsenide wafers today if it were not for government support. The same is true for 4-inch silicon carbide wafers, whose development received substantial government funding that helped drive costs down. He suggested that if the industry had a government partner, it might reach its goals of efficient, effective, low-cost lighting in 5 to 10 years instead of 30 years.
Mr. Trimble reminded participants of Dr. Kennedy’s earlier assertion that innovative applications, such as architecture and office applications, will stimulate the design of different forms of lighting and the evolution from the point source to the panel. At the same time, the availability of these new forms will be important in identifying and opening new markets. These new applications are sufficiently uncertain, however, that startup and even large companies would not be able to sufficiently fund an R&D process. He agreed with Dr. Craford that the industry would eventually move in that direction, driven in part by the economics of energy costs, but the cost of energy would have to rise substantially before it brought a significant incentive to save lighting power.
A Lack of U.S. Strength in Packaging
In response to a question, a participant addressed the challenge of creating larger wafers and appropriate packaging for more powerful LED systems that give off considerable heat. He said that for larger systems, fabrication technologies could be borrowed from silicon or other wafer-processing technologies. He noted that packaging would be a challenge, because thermal management in high-volume, low-cost packaging is not a forte of industry in this country. For optoelectronics, most of the leaders in LED packaging are now found in Southeast Asia and Japan.
Working Toward Higher Efficiency
Roland Haitz gave two suggestions for the most urgent research objectives of precompetitive research. He said that the efficiency in LED-based lamps was likely to be based on a color-mixing system, where a significant problem is to achieve very good color mixing without excessive loss of light. This goal would involve a search through holographic approaches and “whatever else the physics and optical folks can dream up,” and is appropriate for precompetitive research. A second area would make use of Vertical Cavity Surface Emitting Lasers (VCSELS). For this new laser structure, Sandia National Laboratories have recently demonstrated efficiencies above 50 percent at an infrared wavelength of 980 nm. If these results could be extended over the visible spectrum, a VCSEL-based solution would be preferable to the LED-based solution because the photons are easier to direct. Light distribution and extraction efficiency would be substantially more efficient. However, no one yet has demonstrated a VCSEL beyond about 620 nm in the red, which poses a large and interesting challenge for basic researchers.