Semicon West preview: Changing LED manufacturing economics

May 21, 2012
Despite the remarkable recent strides in LED efficiency and lamp quality, the industry still has some ways to go to enable wide adoption of SSL for general illumination writes Paula Doe.
LED and lighting manufacturers must lower costs, improve performance, and probably adopt new business models to make solid-state lighting (SSL) economically viable and spur a mass transition to the technology. The Semiconductor Equipment and Materials International (SEMI) organization has invited industry leaders to talk about potential solutions to key issues of business models, manufacturing yields, packaging costs, and the status of some disruptive technologies in the annual manufacturing technology program at Semicon West, July 11 in San Francisco, CA. This preview touches on some of the most important issues that will be discussed at the conference.

Re-thinking business models to speed adoption

Wide adoption of solid state lighting will require not simply lower costs, but better solutions to reduce risk and increase value for the consumer, argues Mike Watson, Cree senior marketing director, LED components. “Rather than incrementally reducing costs to get to the inflection point to enable the market, it is better to start with the business models that will enable the market,” says Watson, noting that LED lighting is already lower cost initially for most new construction and commercial applications, quite aside from the ongoing energy savings and advantages in instant on, color quality and dimming. But LED lighting still goes into less than about 20% of even new construction, as consumers and lighting specifiers see more risk than benefit. Users are uncomfortable with changing something as basic as their lighting for something of unsure quality and relatively minor savings sometime in the future. Electrical contractors see a threat to their established channels and business. At retail, layers of margin stacking for components and channel markups mean that even if the LED were free, the bulb would still be more expensive than the competition.

To change this risk/return equation for the user, LED makers will need to focus their business on adding value in systems and services that better meet customer needs. This could mean models that reduce upfront cost and better match cost to payback periods, much as mobile phone costs are folded in to service contracts, or as utility rebates for bulb purchases are repaid over time through higher meter rates for electricity (though still lower electrical cost from lower usage). It could also involve supplying services to lighting manufacturers to teach them how to build good lighting products. “We’re still in the stage that is like having to sell semiconductors by teaching people how to assemble circuit boards to make great electronic products,” he says. “You can’t just say here’s an awesome LED.”

Watson notes that auto makers initially had no interest in LED lights because replacing the existing bulb with an LED added to the bill of materials cost. It was only when some started to realize the overall system savings from reducing wiring and metal cutting costs, the added possibilities of using the power saved for other features, and the new design flexibility for product differentiation and branding, that LED sales to auto makers took off.

Changing the economics with GaN on GaN

Another alternative to changing the risk/return trade off for users could come from potentially disruptive technologies like alternative substrate materials. Here Soraa claims early traction in the demanding MR16 replacement bulb market, where it says it offers retailers halogen color quality with enough energy savings for payback in about a year.

Gallium nitride substrates may be costly, but they do eliminate the usual defects and strains from growing GaN active layers on SiC or sapphire substrates, which don’t match the crystal structure or thermal properties of the GaN. That allows some 5 to 10 times more lumens out per unit area, which could potentially change the economics. “GaN substrate costs have come down, since they have been in volume production for making Blu-Ray laser diodes for several years,” says Soraa CTO Mike Krames. “And it allows improvements that mean we can push the LED device to higher performance than with other substrates.” The company figured it could best capitalize on the technology by making high brightness products, starting with a replacement bulb for the 50W MR16 halogen spotlight, a big market with demanding performance requirements for which there was no low energy alternative compact fluorescent available.

Soraa gets higher brightness because it can put in more current without seeing too much droop in light output. Krames explains that the higher quality of the native substrate material enables a design of the active layers that mitigates Auger recombination. But the rest of the process technology also contributes to getting more light out at higher yields and lower production costs. It relies on the significant side surface area of its volumetric emitter to get more light out, as well as to provide a fairly simple process flow of essentially growing active layers, adding contacts and dicing, without thin film issues of substrate removal and thin wafer handling. It cuts the wafer into triangular die to limit internal reflection to extract more light, as once light reflects from the first surface of the triangle it then gets out, without bouncing back and forth as it can between the sides of a square chip. The die are then put on a silicon wafer for wafer-level packaging, which also allows freedom in chip architecture for the microsystem. For the best color quality for its demanding target application, Soraa chose a violet emitter and multiple phosphors including a new down- converting blue, to more closely replicate the composition of natural light, compared to the usual blue emitter with yellow phosphor that Krames says tends to have too little violet, too little cyan, and too much blue.

The company is looking at extending the technology to other high brightness applications, which could range from automotive headlamps and camera flash bulbs to PAR lamps and A lamps. It’s also using GaN substrates for blue and green laser diodes aimed largely at the MEMS micro mirror and pico projector display market.

New materials to bring down the cost of packaging

New materials for package substrates, phosphors and encapsulants have potential to reduce the cost of the LED package, which now accounts for as much as 50%-60% of system cost for a typical 60W replacement lightbulb. “There are ways to reduce the cost of every component by a few percentage points that can add up to a substantial improvement,” says Ilkan Cokgor, VP of marketing at Everlight International in Taiwan. Better thermal plastics look like they can replace expensive ceramic substrates while still maintaining similar reliability, and new production approaches will waste much less of the still relatively expensive improved resins for further savings. Coatings to protect the phosphor from moisture, and improved encapsulation materials can make lower cost phosphors as reliable as more expensive versions. “The final solution would be to eliminate the package entirely, by mounting the chip directly on the heatsink without a substrate,” he notes. “Though reliability and uniform phosphor application remain issues.”

Improving yields by automated inspection and feedback in more steps

Larger wafers and increasingly complex front-end wafer processing are driving LED makers to use more automated inspection and yield management software to improve yields, reports KLA-Tencor. “Automated high throughput inspection of substrates and epi layers for micropits and cracks is already in high volume production,” says Srini Vedula, director of marketing for the growth and emerging markets (GEM) Candela division. “And we’re seeing more interest in automated inspection of patterned wafers now for trying to figure out what impacts yield.” This is driven by the front end processing becoming more complex, with more complex interconnections among defects, especially on high end backlight and SSL products where the cost of yield is very high.

Vedula figures that epi and post-epi patterning each contribute about equally to yield losses, and each potentially have headroom for about 5% improvement in good die out.

“We believe yields are a major key to reducing costs, so the question is how soon the sector can get in-bin yields up enough to reach the inflection point for wide adoption of solid state lighting,” says Frank Burkeen, general manager of the Candela division, GEM group. How well the industry can control yields will ultimately determine the basics of its manufacturing structure, whether arrays of mulitiple small die, or one large die per light source, he suggests, as large die made on large wafers with well controlled in-bin yields could be substantially cheaper at volume than sorting through billions of die for SSL arrays.

Getting more light out with nano patterned substrates

Patterning the sapphire substrate can let out significantly more light, but a lower cost way to make smaller patterns would be more effective. EVGroup suggests both nano imprint technology and a new diffractive process for proximity printing could be efficient solutions.

Patterned sapphire substrates reduce dislocation density to improve internal quantum efficiency, and reduce internal reflection to better scatter the light, on lateral LEDs where the sapphire remains part of the final device. But etching and then overgrowth of these micron-scale pyramidal features is slow and costly.

Nano imprint technology has matured to the point where it can be a viable alternative, says Thomas Uhrmann, business development manager, EVGroup, noting that some companies are using it for photonic crystals or extraction features on top of the LED, and many more have projects in the research stage. But a new alternative that avoids any contact between the mask and the wafer uses a diffraction process that can reportedly print down to 100nm features in periodic patterns with a non-contact optical proximity process, with standard optical proximity resists and low cost mask aligners instead of steppers. The process takes advantage of the Talbot effect’s reproduction of the pattern of light refracting through a grating at a fixed distance away from the grating plane. This reportedly solves depth of focus issues and allows printing on topography without losing resolution. “This means you can change the thickness of the resist and the feature size and pitch by dosage and timing, so you have a bigger playground,” says Uhrmann. EVG worked with Oxford Instruments to demonstrate etching substrate patterns with this system, in a joint project with Eulitha of Switzerland.

These and other speakers at SEMICON West’s LED manufacturing technology program will discuss the how the industry can accelerate the adoption of solid state lighting with both disruptive approaches and more incremental improvements in yields, July 11 in San Francisco.