LEDs: the real story
Griff Resor, Resor Associates, Maynard, MA USA
The hype surrounding a new technology can hide the real story. It is hard to know what's real and what's not. Bad investments can be made or good opportunities can be overlooked. Light emitting diodes (LEDs) are getting a lot of hype. What is the real story? Having lived through the ramp to success for ICs and FPDs, it is clear what steps are needed for LED success.
First, the technology must be ready. After at least 30 years of experimentation, LED technology is ready. The high-volume solution uses GaN materials. GaN diodes emit blue light and drive phosphors to provide other colors. Not all phosphor choices have been settled, but competitive technical solutions exist. LEDs now provide the full rainbow of colors more efficiently than any other lighting technology.
Second, there must be "stair step" markets that pull the technology into steadily larger markets. For ICs these stair step markets were defense, mainframes, industrial computers, personal computers, and mobile products. For flat panel displays, the stair step markets were notebook computers, desktop monitors, and TV. For LEDs, the stair step markets are mobile products, LCD-TV backlights, and solid-state lighting (SSL). Even within SSL there are stair step markets. High-value, lower unit volume applications such as street lights and retail spotlights will come before lower value [per LED], higher unit volume markets in home lighting and office lighting.
Continuous cost reduction is the key. Each stair step market pays for the learning needed to reduce costs. Resor Associates has built a detailed cost-to-manufacture model for the LED process. From this model it is clear that significant opportunities for cost reduction remain. A few examples are presented below.
The key measure of productivity for an LED is cost/lumen. This can be improved by lowering costs, or by increasing light output. The largest cost today is the LED package. Over the past two years, package design has been greatly simplified to lower material consumption, but significant material costs remain.
Gold is a major cost in LEDs. Gold is used for conductor patterns on LED chips and packages, and for wire bonding. With gold now a speculative hedge to economic chaos, there will be a big push to find less expensive materials. Already, flip-chip structures that eliminate wire bonds have appeared.
LEDs are increasing in area, from an average size of 0.5mm2 to 1.0mm2. Two-X larger chips provide just more than a 2X increase in light with very little change in package size or cost. Larger LED chips are clearly an important step towards lower cost/lumen.
To make larger LEDs cost effective, better yields are needed. Yield drops exponentially as chip area increases. Today, yield for 1.0 mm2 LEDs is around 25%. Clearly, significant savings can be made by using cleaner processes and better materials.
Today's LED fab costs about $100 million. Half goes for MOCVD reactors to grow the light emitting layers. Veeco has published its roadmap for improving MOCVD productivity by 3X.
These examples illustrate that the 10X cost reduction called for in the US DOE Roadmap for LEDs can be achieved by making incremental improvements in package design, reducing gold use, increasing LED size, improving yield, and improving MOCVD productivity. These improvements will be realized in time to capture SSL markets step by step. A bright future for LEDs is the real story.
Solid State Technology, Volume 55, Issue 1, January 2012