Guy--
Science could save us more money--what do you think of this piece in Science two weeks ago:
OPTOELECTRONICS:
Blue Semiconductors Settle on Silicon
Robert F. Service
Science Jan 28 2000: 561-562
Researchers trying to coax light from semiconductors have a case of the blues, but
they couldn't be happier. For the past several years they've managed to cajole
semiconductor devices containing gallium nitride to emit blue light when pumped
with electricity. That has opened a world of possible applications, including
converting some of that blue light to other colors and combining them to make a
chip-sized replacement for the light bulb. But blue semiconductor lights are still too expensive for such uses, in part because
they're grown on expensive substrates such as sapphire. Now a team of U.S. researchers reports progress on a cheaper
alternative.
In the 17 January issue of Applied Physics Letters, Asif Khan and colleagues at the University of South Carolina, Columbia,
along with co-workers at Wright Patterson Air Force Base in Ohio and Sensor Electronic Technology Inc. in Troy, New York,
describe a scheme for producing blue and green light-emitting diodes (LEDs) on a base of silicon, the cheap and ubiquitous
substrate for microelectronics. They've also managed to make small LEDs just where they want them on the chip, an important
step toward complex displays made up of thousands of separately controlled lighting elements. The new LEDs atop silicon aren't
yet as bright as those grown on sapphire. Still, "it's a good development," says Fred Schubert, an electrical engineer at Boston
University in Massachusetts. "Silicon substrates are cheap and big. So if a silicon LED technology succeeds, it would mean the
technology could be very cheap."
The garden-variety light bulb, little changed over the last century, costs just pennies to produce but is expensive to run. It pushes
electricity through a tungsten filament, turning it white hot and producing 15 lumens/watt of soft white light and a lot of waste heat.
Newer compact fluorescents, which excite gases to emit light, do better, turning out 60 lumens/watt. But LEDs, which inject
energetic electrons into a solid semiconductor, have the potential to put them all in the shade. As these charges move through, they
shed some of their excess energy as photons of light, the color of which depends on the exact combination of materials used in the
device. And because this process generates far less heat, it can theoretically produce 250 lumens/watt, says Schubert.
But so far reality has fallen far short of theory. Billions of tiny cracks and other defects form in the gallium nitride as it is grown,
and they resist the flow of electrical charges through the device, generating heat instead of light. To minimize the number of
cracks, researchers grow their gallium nitride devices atop sapphire in part because it has a roughly similar crystal structure,
making it easier for the gallium nitride to form an orderly lattice. Silicon's lattice is slightly different, and it suffers from other
drawbacks as well. Most notably, at the temperatures usually used to vaporize gallium nitride and deposit it on the
substrate--around 1000§C--silicon atoms evaporate off and get mixed up in the gallium nitride lattice, causing more defects that
mar its optical properties.
Still, the lure of silicon's low cost has kept researchers searching for a way to make blue LEDs work. In 1998, an IBM team made
some initial progress using a growth technique known as molecular beam epitaxy that works at a relatively cool 750§C. This
produced LEDs that turned out ultraviolet and violet light, albeit about 1/15th the brightness achieved by devices grown on
sapphire. Researchers from the New Jersey-based Emcore Corp. improved matters last September by using a technique known
as metal-organic chemical vapor deposition to lay down a 20-nanometer-thick layer of buffering material on the silicon substrate
and then grow the gallium nitride on top.
Khan and his colleagues combined the two approaches. First, they used a 700§C epitaxy technique to lay down a buffer layer of
aluminum nitride. They then raised the temperature to 900§C and used vapor deposition to create gallium nitride. They also went
beyond the other teams and used the masks and etching of conventional photolithography to place the gallium nitride only where
they wanted it. The result was an array of tiny blue LEDs.
The new LEDs still only put out about one-fifth the light of those grown on sapphire. But Khan and others say they have other
ideas up their sleeve to improve efficiency. The patterning technique also opens up the possibility of making full-color gallium
nitride LED displays. Khan says that his team has already managed to make green LEDs simply by adding a little indium to the
gallium nitride mix. If the South Carolina team or their competitors can figure out a way to also get red gallium nitride LEDs, it
would allow them to integrate both the light emitters and the electronics needed to drive them on the same silicon substrate, which
would drastically drop their cost to produce. That promise is enough to keep the lights burning late into the night at semiconductor
labs around the globe.
Does CREE have to worry about the boys south of the border coming up with a significantly cheaper substrate (i.e., Si rather than SiC)? It sounds like their process is years away from any significant scale-up (I hope) |
