Silicon Is Key to Quest for $5 LED Lightbulb
Bridgelux process grows gallium-nitride on high-volume silicon wafers
Wider wafer: Two-hundred-millimeter silicon wafers make cheaper LEDs.
The LED lightbulb has loads to recommend it. Compared to the compact fluorescent, it can be twice as efficient, lasts far longer, and is free of mercury. But high prices are holding back sales: A 40-watt-equivalent LED bulb with a good hue starts at around US $20, and 60-W versions retail for far more.
The good news is that this barrier to mass adoption should fall in the next two to three years, thanks to recent developments by the LED maker Bridgelux that should spur the launch of a $5 bulb. This California-based firm plans to slash the price of white emitting chips—which account for up to 70 percent of the cost of this type of bulb—by churning out millions of gallium nitride LEDs on 200-millimeter-diameter silicon wafers.
LEDs are usually made on sapphire or silicon carbide substrates that are typically 50 mm, 75 mm, or 100 mm across. Switching to 200-mm silicon would trim the LEDs' cost of materials, because such wafers are cheap and common. However, Bridgelux will realize even bigger savings by outsourcing the processing of its gallium-nitride-on-silicon wafers to the many underutilized 200-mm fabs around the world, says vice president of marketing Jason Posselt. He claims that this outsourcing will cut chip-manufacturing costs by 75 percent and, combined with steady improvements in manufacturing, lead to a $5 LED bulb as soon as 2014.
Being able to manufacture LEDs on readily available, low-cost substrates has been a long-standing dream, according to Ulrich Steegmueller, senior director of advanced development at the LED manufacturer Osram Opto Semiconductors. He and Colin Humphreys, head of the Cambridge Centre for Gallium Nitride, in England, say that most leading LED manufacturers are trying to develop processes to grow their devices on silicon.
"The results of Bridgelux will certainly further encourage this move," says Humphreys. "It is my belief that in the future, all gallium nitride LEDs will be grown on large-area silicon."
Efforts to develop white GaN LEDs on silicon date back 10 years. However, until now these techniques worked only on smaller wafers, and efficiencies have significantly lagged those of commercial devices made on other substrates.
Bridgelux claims it has narrowed this efficiency gap, developing cool-white LEDs on 200-mm silicon that deliver 135 lumens per watt and warm-white variants with a similar color to that of incandescents emitting 85 lm/W. "They are pretty good bread-and-butter numbers, where in [LED] lighting a lot of products are being sold today," says Posselt.
Depositing high-quality gallium nitride layers on silicon is tough; strain results from a large difference in the atomic spacing of the two crystals. Compounding this is a difference in the rate at which the two materials expand and contract when heated and cooled. This may lead to stresses as the gallium nitride film, grown at 1000 ºC, cools to room temperature. If these stresses and strains are not dealt with as the GaN crystal grows on the silicon, the GaN degrades, limiting light emission and driving down efficiency.
"We have found some unique ways to manage that strain," says Posselt. But he would not elaborate on them.
What Posselt will discuss are some of the electronic advantages silicon brings. Bulb manufacturers want to drive more and more current through the LED chips so that they can use fewer of them. However, this leads to lower efficiencies, pushing up operating costs. According to Posselt, silicon-grown LEDs can help, as devices grown on silicon require a smaller voltage increase to boost the drive current through the LED (from 350 to 1000 milliamperes) than do devices grown on other substrates. In addition, silicon is better at conducting heat than sapphire, allowing chips to run cooler and brighter.
Bridgelux will spend the next few years improving yields, qualifying the process, and getting it ready for high-volume production. "It's more development than invention," Posselt says.