Cheaper LEDs Possible by Growing Gallium Nitride on Silicon
Engineers take a step toward cheaper solid-state lighting
5 August 2008—A new method of growing light-emitting diodes (LEDs) on bases of silicon could pave the way for cheaper LED lighting, researchers at Purdue University say. Timothy Sands, a professor of materials engineering and electrical and computer engineering, and his colleagues say they’ve come up with a better way to grow LEDs that are based on gallium nitride (GaN)—essential to white lighting—on silicon instead of on sapphire or silicon carbide, as is often done today. The researchers described the technique in a recent issue of Applied Physics Letters .
Sands’s team were working on a technique called nanoheteroepitaxy, designed to cut down on defects caused by growing one kind of crystal—in this case, GaN—on another. Initially, they placed a thin layer of silicon nitride on top of the sapphire and made tiny holes in it, so the GaN grew only on the holes, reducing the odds of a defect spreading from the sapphire to the GaN. But then they realized that the technique would reduce defects with a different substrate. ”We thought of silicon,” Sands says. ”That’s the logical choice, mainly because you can get it in much larger wafers and it’s cheaper, but it also has better thermal conductivity [than sapphire].” The improved thermal conductivity will allow an LED to be driven at higher voltages, thus producing more light.
One problem with silicon is that it doesn’t reflect visible light well, so a percentage of the photons generated in an LED would be wasted. The team overcame this by inserting a layer of reflective zirconium nitride between the GaN and the silicon. Under normal processing conditions, the zirconium nitride would mix with the silicon, so they also added a layer of aluminum nitride to keep the two separate.
The researchers, working on a three-year Department of Energy–funded project, have not yet built an actual LED this way, but Sands expects that the efficiency of converting electricity to light will be comparable to that of existing LEDs. And the ultimate device would be simpler, because there would be no need to add mirrors to send the light in the right direction, he says.
One potential hitch, however, is that silicon and GaN expand and shrink at very different rates when their temperatures change, which could crack the LED. But Sands says that’s an engineering issue that should be relatively easy to fix.
LED experts describe the work as innovative, if not exactly a breakthrough. Eicke Weber, director of the Integrated Materials Laboratory, at the University of California, Berkeley, calls the work ”a very interesting approach to an important topic” but notes that there are other approaches to growing nitrides on silicon, some of which may produce better results. ”The challenge to operate LEDs with good efficiencies in a heterostructure of this kind will still be substantial, but it appears to be doable,” Weber says.
”I’d call it a significant innovation,” says Steven DenBaars, codirector of the Solid-State Lighting Center, at the University of California, Santa Barbara. He says reducing cost while increasing efficiency is the key to LEDs’ surpassing incandescent and fluorescent bulbs for general lighting, which he believes is inevitable in a few years. But people are working on making larger, cheaper wafers of sapphire and silicon carbide, and DenBaars’s own group is studying bulk GaN as a possible base layer for LEDs. ”It remains to be seen what the best substrate is for solid-state lighting,” DenBaars says.
About the Author
Neil Savage writes from Lowell, Mass., about lasers, LEDs, optoelectronics, and other technology. For IEEE Spectrum Online in July, he reported on a new scheme for networking inside multicore chips and supercomputers.