This article is from the June 15 and July 1, 2001, issue of Red Herring magazine.

Three years ago, Japanese researcher Shuji Nakamura did something with gallium nitride (GaN) that the world's top materials scientists had tried -- and failed -- to do for decades. At a time when most researchers had pretty much written off the difficult, defect-ridden compound, Mr. Nakamura showed that, flaws and all, GaN could emit light brightly and dependably in gorgeous hues of green, blue, and violet.

Engineers have long sought green and blue light emission because, along with the red they already have, green and blue allow them to make bright, full-color displays and efficient, long-lasting white lights. Within a year or two, blue GaN lasers will also be at the heart of improved CD and DVD players and laser printers. Already, sales of GaN devices -- which are used in green LED traffic lights and in thousands of video billboards -- have surpassed half a billion dollars a year, according to the market research firm Strategies Unlimited.

Since Mr. Nakamura's breakthrough, semiconductor researchers have taken a second look at this notoriously frustrating material. Now, a group of scrappy startups, a handful of teams at several high-tech giants, and a few military research agencies are betting that GaN will have as big an impact on power electronics as it is having on optoelectronics. The researchers are testing GaN supertransistors that can withstand extreme heat and amplify signals at frequencies and power levels way beyond those possible with king silicon or any of the pretenders to the throne: gallium arsenide, silicon carbide, and indium phosphide.

TAKE A GANDER If they can find a way to fabricate the transistors economically -- and many are confident they can -- GaN could be the cornerstone of several communications technologies whose widespread implementation depends on far better amplifiers than can be built with silicon transistors. GaN will also likely find a niche in state-of-the-art military radars and communications gear. Further out, rugged, heat-proof GaN transistors might be used to control engines and even electric-power distribution grids. If all goes well, GaN power semiconductors could be a $436 million market annually by 2009, Strategies Unlimited predicts.

GaN developers are looking at cellular base stations as their first opportunity. The amplifiers currently used in these stations are only about 10 percent efficient, which means 90 percent of the power that goes in is wasted as heat. "It's an oven that produces a little bit of wireless signal," quips Pierre Wiltzius, director of semiconductor physics research at Lucent Technologies's (NYSE: LU) Bell Laboratories. The base stations, which are about as big as a standard home refrigerator, are a maze of fans and complex circuitry that prop up as many as 320 silicon transistors that are being pushed to their limits.

Those beleaguered silicon transistors probably won't be sufficient to handle reliably the data volumes anticipated with the arrival of third-generation wireless. Mr. Wiltzius says GaN transistors could boost base-station amplifier efficiency to 20 or 30 percent and "transform the way people do wireless. If you could shrink a wireless base station to the size of a dormitory refrigerator, all of a sudden you could hang it on top of a telephone pole rather than having to rent space in a central office."

The conventional wisdom is that GaN transistors won't be available commercially for at least four or five years. But Nitronex, a Raleigh, North Carolina, startup says it has made a technological leap that could cut that timetable down to one or two years. The company has already delivered GaN transistors to potential customers for testing; the group reportedly includes Ericsson (Nasdaq: ERICY) and Motorola (NYSE: MOT). Nitronex plans to start selling GaN transistors next year, according to its president and CEO, Bob Lynch. Last November the 45-employee company closed on $9.5 million in financing, with major investments from VantagePoint Venture Partners, Alliance Technology Ventures, Southeast Interactive Technology Funds, and others. John Marren, a partner in the Texas Pacific Group, and David Norbury, president and CEO of RF Micro Devices, also invested.

BEST OF BOTH WORLDS Three attributes make GaN the closest thing so far to an ideal semiconductor material. First, its high "electron mobility" and "saturation velocity" -- two qualities related to the speed at which electrons zoom through the material -- mean it can amplify high-frequency signals. Second, GaN can withstand high voltages without breaking down and conducting electricity -- very important for amplifying signals without distorting them. Third, when electrons combine in GaN with electron deficiencies called "holes," the semiconductor releases larger amounts of energy than do other materials when their electrons and holes get together.

So, given all its amazing properties, why aren't GaN transistors everywhere? Because there's currently no cheap substrate material for GaN. A substrate is the semiconductor's foundation. It starts out as a wafer on which the transistors or circuits are grown. To prevent performance-robbing structural flaws between the substrate and the semiconductor, engineers usually use the same material for both. But no one has yet figured out how to grow GaN into the large crystals from which the wafers are sliced, salami-style.

Engineers are now using sapphire and silicon carbide as the substrate for GaN devices. The problem is that these materials are expensive. Two companies -- RF Nitro Communications of Charlotte, North Carolina, and ATMI (Nasdaq: ATMI) of Phoenix, Arizona -- make GaN-coated wafers of sapphire or silicon carbide that can be turned into transistors. A 2-inch sapphire wafer costs about $2,000; the same-size silicon carbide wafer runs as high as $12,000. (Roughly 270 transistors can be grown on a 2-inch wafer.) Labs at DaimlerChrysler (NYSE: DCX), General Electric (NYSE: GE), NEC Electronics, and Raytheon (NYSE: RTN) are making and testing transistors on RF Nitro's wafers.

COMPOUND INTEREST Meanwhile, Nitronex is growing GaN on cheap (under $50) silicon wafers. Its secret is a method the company calls Pendeo, a variation on a technique called lateral overgrowth. Nitronex's technicians first grow a layer of GaN on the silicon wafer, then a layer of silicon dioxide, followed by another of GaN. The layer of silicon dioxide blocks many flaws in the substrate from intruding into the top layer of GaN, where transistors can then be fabricated. The concept is attractive, but so far Nitronex has not revealed how well the resulting transistors work.

Other companies are in pursuit of a more ambitious goal -- wafers of pure GaN on which perfect layers of the semiconductor can be grown. These firms include Astralux of Boulder, Colorado; Technologies and Devices International in Gaithersburg, Maryland; Crystal IS in Latham, New York; and Kyma Technologies in Raleigh. Kyma expects to begin testing 2- and 4-inch GaN wafers later this year.

For now, the GaN supertransistor is an expensive technology in search of a killer app. Better semiconductors for cellular base stations are one possibility, but that entire market is worth only about $261 million a year, according to Strategies Unlimited. Another prospect is the amplifiers within the modulators in fiber-optic systems -- these convert electrical signals to light and vice versa. When they increase from 10 to 40 Gbps in the near future, as expected, the amplifiers will have to swing 15 to 20 volts at 40 GHz. "You cannot do this with silicon, gallium arsenide, indium phosphide, or anything else," insists Lester Eastman, a leading GaN researcher at Cornell University. "But we think we can do it with gallium nitride." He says he has already tested GaN transistors at an astounding 140 GHz.

Asked if modulators for fiber optics is one of the applications Bell Labs envisions for its GaN work, the normally voluble Federico Capasso, director of the Physical Research Laboratory, just smiles. "I can't tell you about it," he says. "The level of competition with the rest of the world is incredible."

Just what the mystery material needed -- a little more intrigue.

Write to letters@redherring.com.

The silicon killer

Explained: The holy wafer

– Full circuit

Marriage of expedience

Fabricating photons

Image is everything