Update from Strategies in Light:
HB-LED Focus Turns to Cost Reduction and Next Generation Digital Intelligence
By Paula Doe, SEMI Adjacent Markets
High-brightness LED technology has made impressive progress in brightness and efficiency over the past few years, but big penetration into the general lighting market still requires approximately a 10X reduction in costs to compete with fluorescents, which are also significantly improving. That’s creating a window of opportunity for both improvements in volume manufacturing with current materials and equipment, and for disruptive new approaches.
Sales of packaged HB-LED devices jumped 93 percent in 2010, to reach $10.8 billion, according to Strategies Unlimited, as continuing improvements in efficiency, uniform color, and reliability made the chips an irresistible choice for displays with thin designs, better color and low energy usage. Display backlights account for the majority of the market for HB LEDs — some 62 percent or $6.7 billion — and will continue to drive sector growth for the next couple of years as the giant television market continues to convert to LEDs. The next wave of growth will come from general lighting, likely starting to accelerate around 2013-2015, when governments around the world start to ban the inefficient incandescent light bulb — assuming, that is, that costs do in fact plunge by an order of magnitude.
The major HB-LED makers are reporting that device technology is now good enough for the more demanding general lighting applications. Devices that produce more than 100 lumens per Watt, and average more than 300 lumens per 1W package, are commonly available. This makes it possible to make LED lights as bright as incandescent 40W and 60W bulbs, after the losses from the package and the system, while using less electricity than competing technologies. “The LED is no longer the gating factor for solid state lighting,” argued Ella Shum, director of LED research for Strategies Unlimited.
Aldo Kamper, president and CEO of OSRAM Opto Semiconductors concurred that LED growth would come against all types of conventional lighting technologies because it can now compete on efficiency in all applications. He also suggested that further improvement in efficiency was becoming more difficult, and also less necessary, noting that the economical limit of efficiency for warm white may be ~160lm/W. Michael Watson, senior director of product marketing and applications for Cree Inc., also argued that efforts to improve efficiency may be reaching the point of diminishing returns.
Major progress has also been made recently in consistent white color. Though tool suppliers are still selling sorters that put die into 120 or 130 separate bins for color, forward voltage and brightness, leading companies like Philips and Cree are now matching phosphor to die or die array at operating temperatures of 85°C, to make packaged white LEDs consistent within one 3- MacAdam ellipse bin— a level of variation barely detectable to most people.
Focus Moves to Cost Reduction
That moves the focus to cost reduction. Despite the bad reputation of compact fluorescent bulbs, fluorescents do dominate the lighting market worldwide, with their universal use in commercial and industrial applications, and common residential use in Asia, and that technology is low cost, efficient, and improving. “The industry has matured enough so that it’s now all about costs,” said Dan Coyne, managing director, Canaccord Genuity.
Both incremental and disruptive solutions will likely be needed. Osram’s Kamper said he counted particularly on moving to larger diameter wafers, improving the homogeneity of the epi layer, and improving yields.
Michael Holt, CEO of Philips Lumileds Lighting Company, suggested that there was particular potential for improvement in phosphors, and in thermal management of the systems. Though the overall margin for improvement in epi may be limited, he argued, there is still room to improve the lumen output per wafer area to reduce die size, by better marriage of the epi to the substrate and by reducing the droop in performance at higher power levels. For the droop issue, he expected improvement, if not solution. “Someone will move that curve out so the sweet spot is closer to 3A/cm2,” he said. “And there are lots of ideas for better systems. It will be a long time until we’re in a commodity war.”
Cree’s approach to cost reduction has been to design the system for ease of manufacturability, and for ease of use by customers. Watson argued the key was designing for manufacturing throughput with fewer optical sources, and by color binning by design— testing and binning at operating temperature to get 2 or 4 step consistency. Cree is also working with partners so they can develop appropriate drivers, optics and thermal solutions designed for best performance with a particular HB-LED device from the beginning. Users can then save on costs of supply chain management and assembly.
Another option where investors see opportunity is in reorganizing the value chain to reduce the margin stacking that now drives up prices, from so many different players involved in making and selling the final lighting fixture. “What parts of the value chain to best put together? Where will the margin be?” questioned Jed Dorsheimer, managing director of equity research at Canaccord Genuity.
Options to Improve Manufacturing Efficiency
At the manufacturing workshop SEMI sponsored at the conference, member companies talked about the specific technologies to improve yields and reduce costs, particularly stressing the gains from more sophisticated metrology early in the process flow.
Veeco chief technologist Bill Quinn noted progress on more precise control of the temperature of epi deposition, thanks to the development of a UV pyrometer with Sandia Labs that can measure the temperature of the transparent wafer directly, instead of just that of the carrier. Yields can also be improved by onboard tool diagnostics, with process control software to analyze data in real time and make adjustments, as well as close-coupled metrology, such as optical mapping, right after processing. With the critical device wavelength changing by ~2-3nm for every degree C variation in a 750°C process, and for every angstrom variance in film thickness, tight control is vital. “You need to catch yield losses early,” said Quinn. “Or you start with junk and end up with very expensive junk.”
Rudolph Technologies Inc. noted that device makers were moving from go/no go tests to adding process control through the line, but had yet to add the control systems to turn that data into decisions. VP and GM Mike Plisinski argued that automated fault detection that aligned wafer maps across processes and tools, identified spatial patterns of defects, and then correlated those defects with yields could identify the causes of the defects that mattered. He said customers reported that the automated generation of diagnostics instead of just alarms saved 2-4 hours per day per engineer by allowing them to get straight to solving the problem.
Another step towards cost reduction was taken by 30-some senior industry technical folks who got down to work on ways to insure interoperability on the recently formed SEMI LED task forces on wafers, automation and assembly standards, aiming to ease the way to enabling the automation needed for high-volume manufacturing for solid state lighting. First efforts are focusing on reaching consensus on placement of flats and notches on 6-inch sapphire wafers, and common cassettes and software and hardware interfaces. “I’ve never seen standards kill innovation,” said committee member Chris Moore, president and CEO of metrology supplier Semilab AMS, in response to a question from a skeptic. “It’s all about making the process plug and play, so what you buy works with what else you have.” The HB-LED Standards Committee and task forces will next meet at the end of March at SEMI’s offices in San Jose. For more information, visit http://www.semi.org/en/Standards/ctr_042107.
Potentially Disruptive Solutions for Substrates and Phosphors
Substrates and phosphors probably provide the most opportunity for disruptive cost saving technology, and companies are reporting progress on both fronts. LatticePower CTO Hanmin Zhao, reported reliability tests on the Chinese company’s GaN on silicon LEDs that claimed similar performance and degradation as sapphire products from a leading supplier. The company makes its thin-film, flip-chip style devices by selective growth on mesas on pre-patterned 2-inch silicon substrates, growing 3-4 microns of GaN on top of a superlattice transition layer for a 5-micron thick film, using multiple quantum wells to maximize efficiency, and removing the silicon substrate by low cost wet-etching instead of having to do laser liftoff. Zhao said the company gets about the same performance on silicon as it does from the same epi layer stack on sapphire, reporting a run of >1000 chips yielded 55 percent within the range of 447.5-462.5nm wavelength, with Vf between 3.0-3.4V, and light output power >300mW. Wafer maps showed most of the defects on the edge. It’s running 100 2-inch wafers a day now, selling a 960µm bare chip, but has also invested in a packaging company, and has a sister luminaire company that is testing the LEDs in street lights.
The German epi foundry Azzurro reported its GaN on silicon process is growing >5µm GaN stacks on 6-inch wafers, while keeping the wafer bow under 20µm so it can be processed with conventional CMOS tools. Alexander Loesing, CFO, VP Sales and co-founder, said the company was getting brightness comparable to sapphire and similar dislocation density. It targets moving the process to 8” wafers by the end of this year, expanding the capacity of its pilot production line. The GaN on silicon epi wafers also claim breakdown voltages of >750V for high voltage power semiconductors.
Some disruptive solutions on the phosphor side are a little further along, in design for commercial products targeted for release later this year. Nanosys Inc. is working with Samsung and LG to ramp production of its quantum dot alternative to phosphors for edgelights for mobile device displays. CFO John Page said red and green quantum dots in a glass tube in front of the blue LEDs produces narrow spectrum red, green and blue light, for brighter, more saturated color, with 20% better efficiency, and at lower cost than YAG. The company raised an additional $31 million in October, in a round led by Samsung.
Meanwhile phosphor maker Intematix is printing a carefully tuned blend of multiple phosphors on transparent glass or plastic sheets, using a proprietary hybrid printing process, for consistent remote phosphors that also replace the diffuser optics. Chuck Edwards, VP of product development, says this system-level alternative to package-level phosphor coatings can boost system efficiency by as much 30 percent, and reduce costs at the system level by easing requirements for the power supply and heat sink. The phosphor discs can be matched to the average wavelength of an array of blue chips to limit binning. He says the company is in the design-in phase with several beta customers.
What Comes after Lightbulb Replacement?
The LED world remains confident that bringing semiconductor technology to the lighting industry will create new markets for intelligent lighting that will keep the HB-LED market on a steady growth path even after most light bulbs are replaced with LEDs that last for decades. Smart lighting systems that significantly saves energy by automatically adjusting brightness to just the level needed have gotten the most attention, but there are plenty of other ideas as well. Osram’s Kamper ticks off a list that includes adaptive auto headlight systems that automatically dim the brights for oncoming cars and swivel around turns, active infrared night vision systems for autos, LED projectors in the cell phones, optical gesture recognition for gaming systems, high-powered flash bulbs for digital cameras, and better grow lights for green houses. “I don’t know yet what is coming next,” he said. “But it doesn’t worry me, because we’ve developed new markets for new digital technology over and over again.”
Kamper also projected that demand for LED fab capacity would continue to increase even after a theoretical 100 percent penetration of LEDs into general markets. The market will likely be less cyclical than, say, the DRAM market, with more modest swings like the power electronics market, as the high currents limit scaling, products have to be tailored to a wide variety of specific applications, and lower-cost fabs mean capacity can be added in smaller chunks.
UV-LEDs open other possibilities. LEDs would bring the same long life, high efficiency and targeted emission spectra to the UV world, noted Cary Eskow, global director of Avnet LightSpeed. Unlike the current mercury discharge UV lamps, LEDs can switch on and off instantly, expanding the possibilities for UV flash curing and printing. UV lighting can also stimulate production of healthful antioxidants in plants and vitamin D in people, highlight and then zap target cells in biomedical assays, and perhaps even be reflected off the atmosphere for daylight free-space optical communication. Others suggest using the optical frequencies of LEDs instead of the crowded radio frequency bands to transmit data.
Venture capitalists said they were now looking for energy stories with potential to increase revenues or increase functionality, for more upside than just reducing costs. “The interesting stuff is what will happen when the physicists are done with the basics,” noted Bob Walker, partner at Sierra Ventures.
For more information about HB-LED at SEMI, visit www.semi.org/led
March 1, 2011