May 2003

LEDs vs. the Lightbulb

Continued from page 2

By David Talbot

Getting the LEDs Out

 A white LED is only as good as its weakest component. Its performance depends on such factors as the purity of the semiconductor materials, the shape of the phosphor crystals, and even how well the lit device is able to dissipate heat. To bring white LEDs from the lab into broader markets will require improvements-and cost reductions-in all of these elements.

The job starts in a very hot oven. The key ingredient-gallium nitride-is born from gases fed into a superheated chamber. There, molecules containing gallium and nitrogen break apart, and crystals of gallium nitride begin to grow atop a sapphire substrate (in a process akin to the making of computer chips). It takes hours to deposit the dozens of layers, each with a slightly different chemistry. And it is far from a perfect process. A single sample, depending on where it is energized, can produce light of several different colors-which is why Nakamura is poking and prodding his materials so meticulously. Moreover, subtle differences in the arrangement of atoms can result in efficiency-robbing regions that look like vertical tunnels in the material.

Improvement of efficiency calls for growing crystals without those tunnels. Researchers at Sandia National Labs in Albuquerque, NM, believe they have a simple method for doing that, thus producing brighter LEDs. They start by etching grooves into the substrate, leaving a series of thin sapphire ridges, each about one micrometer wide. These ridges act like floor joists. The gallium nitride grows atop the sapphire ridges and moves sideways out over the grooves. Experiments have shown that this method reduces the number of defects by two orders of magnitude-and boosts brightness tenfold.

Even when the material itself is efficient, the light must go where it's needed. Much of the light produced by a typical blue LED bounces around within the structure and is wasted. Nakamura's group at Santa Barbara is adding mirrorlike nanostructures, just 50 nanometers wide, to certain crystalline layers. Steven DenBaars, a materials scientist who works with Nakamura, says that adding the nanostructures increases by 50 percent the light that gets reflected out of the device. But because such layers have not yet been integrated into a finished LED, the full payoff is uncertain.

Size matters too. A big LED chip gives off more light than a small one. Pursuit of a bigger LED has been a focus of the work at Durham, NC-based Cree, which makes some of the brightest blue LEDs on the market and is one of several corporations that have partnered with Nakamura's Santa Barbara lab. Last year Cree introduced an LED chip measuring 900 by 900 micrometers. It provides nine times the light-emitting surface area of the 300- by 300-micrometer chips that had been the industry standard. This expansion yields a simpler, hence cheaper, device, says Cree vice president Norbert Hiller. Lumileds, too, is developing larger chips and was able to deliver one of the world's brightest white-light LEDs: a  five-watt device that puts out as much light as a 10-watt incandescent bulb.

Advances are also needed in phosphors. That's exactly what GELcore-a joint venture of General Electric and Somerset, NJ-based Emcore-says it has achieved. Starting with LEDs that emit ultraviolet light, researchers at GE's Global Research Center in Niskayuna, NY, went to work improving the phosphor recipes developed for the company's ordinary fluorescent light tubes. The result: they increased by a factor of 100 the ability of the phosphor to absorb energy, says Charles Becker, a physicist and manager of LED lighting research at GE's research center.

Thanks to this development, GE says it is close to launching a white-light device that can produce 30 lumens per watt, a considerable improvement over the 10 to 15 lumens per watt rating that consumers are accustomed to seeing on the boxes of typical incandescent bulbs. What's more, the device is designed to last 50,000 hours-about six times the typical incandescent's lifetime. GELcore plans to begin selling LEDs that use this technology later this year and aims to push light output to 50 lumens per watt within two years, says Becker.

Just about every company in the white-light LED game is tackling these and other methods. One approach skips the phosphor conversion process altogether and instead produces white light by combining the output of red, green, and blue LEDs. Lumileds, for instance, is already assembling emitters of these three colors to produce backlighting for displays such as those used in cell phones and laptop computers. While bypassing the phosphor adds cost and difficulty, this technique increases efficiency.

Unfortunately, the materials of each color LED degrade at different rates. Stabilizing the white, therefore, requires sensors and electronics: as the sensor records a decline in an LED's output, it directs the circuit to compensate by feeding more power to the chip. Steranka says that making this approach work for a low-cost general-illumination product calls for volume-driven cost reductions in such electronics, as well as more efficient individual LEDs. But the payoff would be further gains in efficiency-a goal Steranka is confident can be achieved. "People know how to do to it," he says. "It's not rocket science."