In a plenary talk he gave at a conference in Strasbourg, France, in June 1998, Shuji Nakamura made what would turn out to be an overly bold assertion. The essence of what he said was, LED research is over: there is nothing left to do. In the audience that day was Nakamura's good friend, Asif Khan. Afterward, Khan went up to him and said, “You know, Shuji, that's a great challenge—I'll show you that LED research is not finished, that there are still things to do.”

Back at his base at the University of South Carolina, Khan considered what actually remained to be done in nitride-based LEDs. In addition to picking up the gauntlet that Nakamura had thrown down, Khan was also motivated by a second consideration. Namely, that as a professor leading a research group, he could not merely be repeating what others had done. He always had to be thinking what the next frontier would be. In a sense, the answer was obvious. The history of light emitting devices has been, as we have seen, a progressive shortening of the wavelength—from infrared, to red, to yellow, to green, to blue and violet—and a corresponding increase in the bandgap, from narrow to wide. That left ultraviolet, “black light” as it is sometimes called, because UV is invisible to the human eye. (But, as any nightclub goer knows, ultraviolet light causes white clothes and teeth to glow in the dark.)

Ultraviolet light subdivides into several categories. There is UV-A, wavelength 380–315 nanometers, also known as long wave; UV-B, 315–280 nanometers, aka medium wave; and UV-C, below 280 nanometers, aka short wave. Big LED companies like Nichia and Toyoda Gosei had focused on making gallium nitride devices that emitted wavelengths at around 380 nm. These LEDs could pump a phosphor that converted ultraviolet to white light. They could even, as Toyoda Gosei had quietly shown, be used in conjunction with a photocatalyst to produce air purifiers as an option for Japanese luxury cars. But thus far no one had ventured below 380 nm, into deep UV territory.

The reason was simple. To get to shorter wavelengths meant moving to a different alloy, aluminum gallium nitride. And, as Khan knew well from his experience, AlGaN was extremely difficult to grow. Add aluminum to the mix and it was a different story, complete with its own distinctive set of technological barriers. Aluminum nitride itself is an insulator, not a promising starter material for devices that rely on the conduction of electric current. In addition to which, the more aluminum you add, the more cracks in your material you get. Not good because, to widen the bandgap sufficiently to reach deep UV, the light emitting layers of devices would need to contain a great deal of aluminum. On the other hand, if you could actually make aluminum gallium nitride LEDs, then a whole new world of applications would open up. In particular, deep UV LEDs had the potential to replace the conventional source of ultraviolet light, mercury lamps. Semiconductor light emitters would, as usual, be smaller, would draw less power, and last much longer. They would also, in this case, be environmentally friendly, unlike mercury, which is a potent neurotoxin (not a nice thing to have leaching from landfills into the water supply). Even at very low levels, mercury can damage the central nervous system, the liver, and the kidneys. Mercury is especially hazardous to pregnant women and young children.

Asif Khan was, he felt, intuitively good at figuring out areas that would be fruitful to pursue. He had after all picked up gallium nitride back in the early 1980s when no one else in the United States was working on it. He and his group had been the first to demonstrate that it was possible to make quantum wells out of the nitrides. But he had been way ahead of his time. Through a combination of bad timing, bad management, and bad luck, in the race to make visible gallium nitride light emitters, Khan had lost out to Nakamura. Now, his instinct told him that deep ultraviolet LEDs were going to be big. “I'm going to go with my gut feeling that this will be an important area,” he told his students. “Sometimes you have to take a gamble, you know?”

Khan would bet on deep UV LEDs. In this research, ironically, he would wind up competing with Nakamura and his group at University of California at Santa Barbara. But this time, Khan would win.

In some ways, the achievements of Khan's group at the University of South Carolina mirror those of Nakamura at Nichia. Here, too, modification of the way gases are introduced into the MOCVD reactor jar was crucial. Modification avoided the formation of film-destroying chemical “snow,” aluminum being especially eager to react with ammonia. Also vital was the ability to come up with recipes for complex device structures, for depositing stacks of quantum wells that would shepherd recalcitrant electrons and holes to their light-producing destinations. Happily, the group included Wenhong Sun, a world-class crystal grower with a Nakamura-like ability to bend MOCVD systems to his will.

After plugging away at the problem for three years, in mid-2001 Khan's group finally made a breakthrough—an LED that produced appreciable amounts of ultraviolet light at 340 nanometers. They had demonstrated that UV LEDs were feasible. Having made something that worked at UV-A wavelengths, over the next few years the USC researchers pushed on, modifying their approach, retuning their reactors to develop devices that emitted UV-B and UV-C light at wavelengths all the way down to 250 nanometers.

The breakthrough occurred at a propitious moment. Shortly afterward Khan happened to be visiting the headquarters of the Defense Advanced Projects Research Agency at the Pentagon. In a corridor—the Pentagon is mostly corridor—he happened to bump into John Carrano, one of the agency's bright young program managers, whom he had previously met at a nitride workshop. Carrano asked Khan what he was up to. Khan told him about his exciting UV LED results. “John said, Perfect timing, because I'm about to start a large program on deep UV light emitters for biological agent detection, and I would love you guys to participate.”

DARPA's Semiconductor Ultraviolet Optical Sources (SUVOS) program would run for five years, with funding of around $50 million. Seventeen projects would be funded, with Khan's group at South Carolina, Nakamura's group at Santa Barbara, and Cree among the principal recipients of Department of Defense largesse. For this program, Khan hired Herb Maruska, who had grown the first GaN films at RCA Labs back in the 1960s. Maruska designed and built a hybrid reactor that the group at USC would use to grow thick films of AlN and AlGaN to provide substrates for ultraviolet LEDs.

The military had three reasons to be interested in black light. One is that deep ultraviolet light is germicidal. It kills viruses and bacteria like E. coli (and, were there not a protective ozone layer to absorb it, ultraviolet radiation from the sun would kill us, too). This means that UV light can be used to sterilize water and make it drinkable, a handy trick for troops out in the field. It is not practical to carry mercury lamps, which are bulky and fragile, into battle. But a swizzle stick tipped with a solar-powered UV LED would fit into any soldier's backpack. A company called Hydro-Photon based in Blue Hill, Maine, is already working on a commercial version of such a portable sterilization device.

Covert communications were a second area of interest. Since essentially no deep UV light from the sun reaches the earth, “solar-blind” wavelengths provide an excellent, interference-free window for certain applications. These include networks of ground-based sensors that can detect activity—like bad guys moving around exclusion zones—even in broad daylight.

But the most important reason for wanting a UV LED from DARPA's point of view was as a sensor capable of detecting biological warfare agents, notably anthrax. This was an application that acquired a sudden urgency in September 2001, with the arrival of letters containing anthrax powder at the offices of several US senators and media outlets. Five people died as a result of exposure to this noxious stuff.

To determine what a substance is, whether harmless particle or deadly poison, you shine ultraviolet light on the sample and measure the resulting fluorescence. Conventionally, this is done using extremely expensive, laboratory-based gas-laser systems. With the advent of deep UV LEDs, it became possible to build a portable biosensor system about the size of a laptop PC, costing a fraction of the price. No wonder, then, that Carrano would describe the results of his SUVOS program as “wildly successful.”

“It won't take you long to find the university,” Warren Weeks had told me, and it didn't. On a sultry October morning I drove into Columbia, South Carolina's pint-sized capital, passing through rundown neighborhoods of two-story, pastel-colored clapboard houses where, it seems, only African-Americans live. A few blocks away from the city center, I arrived at a massive, apparently windowless, triangular concrete pile. It houses the University of South Carolina's Department of Electrical Engineering, which Asif Khan chairs. The man himself turns out to be fifty-five years old, of slight build and medium height, with salt-and-pepper hair and large, hooded brown eyes. His manner is no-nonsense, but friendly and engagingly frank.

Asif Khan was born in Meerut, the former British army cantonment north of New Delhi, where the Indian Mutiny of 1857 began. His father's people were mostly either teachers or doctors; his maternal grandfather ran a business that made brass musical instruments such as flutes and tubas. Khan's father was a statistician at the Indian national weather service. He was posted to Lahore in what, after the Partition of India in 1947, would become Pakistan. Although most of Asif's relatives remained in India, the name Khan is actually Pathan in origin. This suggests that the family must at some stage have migrated down from the northwestern frontier region of Afghanistan.

On graduating from the University of Karachi with a master's degree in nuclear physics, Khan came to the United States, to MIT, where in 1979 he got his PhD. He joined the Honeywell Research Center in Minneapolis, which in those days was one of the top laboratories working on compound semiconductors. As it happened, Honeywell has a big business making industrial furnaces. The pilot flame in such furnaces emits UV light. A UV sensor, with its narrow window of solar-blind sensitivity, can clearly detect whether the pilot flame remains lit.

Honeywell was keen to replace the conventional flame detectors, photomultiplier tubes. These are expensive, fragile, short lived, and require a high voltage to operate. The company threw the problem to its research center. Gallium nitride UV detectors seemed like a possible solution. So, by a twist of fate, when Khan decided to work on UV LEDs at USC, which he joined in 1997, he was returning to the theme that had been his original point of entry into professional research.

Today, Khan heads an organization that is in the forefront of nitride research worldwide. It is an impressive setup, as I discover when Asif takes me across the road for a tour of his pride and joy: the converted three-story warehouse in which are located the laboratories that he founded and directs. As we move from one room packed with high-tech gear to the next, I ask him whether it is difficult to operate down here in Columbia, a long way from the nearest high-tech nexus. “On the contrary, it's an advantage,” Khan tells me. For example, he is not bothered by red tape regarding regulatory issues such as environmental concerns that would be the case were his labs located in the likes of Silicon Valley. Asif has garnered much support from South Carolina's powers-that-be. The local authorities are evidently delighted to have lured a researcher of his caliber to their state. They see him, rightly, as an engenderer, from whose center of excellence good things—prestige, companies, jobs—can be expected to flow.

As we walk from one building to the next, I press Khan on the cultural, as opposed to operational, privations of life in the Palmetto State. Does he miss cricket, a sport to which almost everyone from the subcontinent is passionately devoted? “Not at all,” Asif replies. Though he himself has not played much recently, the game flourishes in the southeastern United States. There is a local league, with each city fielding its own team, traveling from place to place each weekend to play. Ah, the unexpected fruits of globalization!

My tour of the labs ends, leaving me with the feeling that I have seen up-close a great thrumming engine of research and development in action. Driven by a dynamic individual, the engine is powered by a phalanx of applied brainpower, measured in professors (five), grad students (twenty), and postdocs (fifteen), all relentlessly pushing the edges of the technological envelope. Many of the staff are not American-born; in particular, there is a large contingent from Asia. I cannot help wondering how many of them will eventually return to Taiwan, Korea, and China to guide the solid-state revolution in lighting as it takes off there.

One useful metric for the value of the outcomes of a laboratory is whether any start-ups have formed to commercialize its research. As Bob Davis's lab at North Carolina State begat Cree, so it seems Asif Khan's laboratory at South Carolina begat Sensor Electronic Technology. In fact, the story is more complicated than that. SET was actually founded in 1999 in upstate New York, by two professors from Rensselaer Polytechnic Institute, Remis Gaska and Michael Shur. Gaska had previously worked for Khan as a research scientist at APA Optics, an ill-starred start-up that specialized in nitride technology. The original idea was that SET would develop high-performance, nitride-based electronic devices such as transistors and sensors. Khan signed on as a consultant to the fledgling firm.

When they heard that Khan had decided to go after UV LEDs, Gaska and Shur were initially dubious about his chances of success. But Khan kept badgering them about it, showing them his group's results, urging them to switch their focus to UV LEDs, insisting that there was tremendous commercial potential in such devices. Gaska and Shur remained skeptical: they knew the formidable problems involved in fabricating aluminum gallium nitride devices. But Khan's demonstration that UV devices were doable finally convinced them to get on board.

In 2002, having secured funding from DARPA, Gaska and Shur moved all their operations from Troy, New York, to Columbia, South Carolina, in order to have close cooperation with Khan's group. The relationship between start-up and university is close. SET takes advantage of the university's expensive tools to analyze its materials. Khan and his longtime collaborator Jinwei Yang are both part owners and directors of the company. Five of its twenty-one employees are former students and postdocs from USC, including Jianping Zhang, a leading crystal grower.

SET took the USC technology and ran with it, working on the performance of the LEDs, improving their efficiency. In May 2004 the company's researchers made a major change in the device design that enabled them to increase the output power of deep UV LEDs by as much as twenty times. This brought the light up to commercially significant levels. They filed patent applications. By the following August, SET had research-grade sample products ready for market.

It takes about ten minutes to drive from the university, through more dilapidated streets, to the outskirts of Columbia where, off an unkempt side road, the facilities of SET are located. They consist of two aluminum-siding warehouses; more like big sheds, really. An unlikely setting for a high-tech outfit, perhaps, but no more so than that of Nichia, in remote Anan. Outside the company's front door is parked an army green Hummer. The state motto on the behemoth's license plate reads “Smiling Faces, Beautiful Places.” Not a bad vehicle to drive, I reflect, if a large chunk of your funding comes from the US Department of Defense.

The Hummer belongs to Remis Gaska, SET's president and CEO. A tall, hefty, blond fortyish man, he is resigned to people asking him about the origins of his unusual name. Remis is in fact a shortened version of remigijus, which apparently means “rower” in Latin. Born in Lithuania, Gaska moved to the United States in 1992, after the collapse of the Soviet system. He has two PhDs: one, in physics, from Vilnius University; the other, in electrical engineering, from Wayne State University in downtown Detroit.

People from the Baltic states tend to be laconic—life is hard up there—and Gaska is no exception. But there is no mistaking the excitement in his voice as he explains the extraordinary developments that are taking place in UV LEDs. “It's a revolution,” he says. “You'll see!”

Ralph Waldo Emerson famously remarked, “If a man can…make a better mousetrap…though he builds his house in the woods the world will make a beaten path to his door.” He might well have been talking about SET and its range of UV LEDs. As the world's sole supplier of sub-350-nanometer devices, without any advertising or marketing, merely by word of mouth, SET has already attracted more than three hundred customers. Every week, several more beat a path through the woods of South Carolina to the company's door.

Ultraviolet LEDs turn out to have an astonishing variety of applications, including some the company never initially imagined. One difference between the markets for visible and invisible LEDs is that, whereas in the former the exact wavelength of the light does not matter so much—if it's blue, it's blue—in the latter, customers often have very specific requirements. SET is focusing on about a dozen different wavelengths, from 365 nm down to 250 nm. Gaska walks me through the main applications for the devices.

One range is between 365 and 340 nanometers. There, a lot of the interest comes from the curing—rapid drying and hardening—of polymers. These could be inks, or fiber optic coatings, or dental fillings. The small size of UV LEDs is also of interest to the semiconductor industry, to cure the photoresists used to create circuit patterns in microchip lithography. These are all multibillion-dollar markets.

Another range is between 340 and 310 nanometers. This is a prime area of interest for medical applications. Since skin is particularly sensitive to 312–310 nm, dermatologists use UV light to treat skin conditions such as psoriasis and certain allergies. In addition to these existing markets there are also emerging medical markets, such as photodynamic drug therapy, which is used to treat cancer. Once injected, such drugs concentrate in the cancer cells, where they are activated by shining a light on the affected area. Here again the small size of UV LEDs is a major advantage, since it means that biomedical equipment can be miniaturized and made portable for on-the-spot therapy.

A lot of interest in the range between 300 and 295 nanometers comes from the biological research community. As we have seen, bioagents such as proteins are intrinsically fluorescent when excited with UV light. Current lamp- or laser-based optical spectroscopy systems are typically cumbersome and very expensive. A UV-LED-based system could reduce the cost by an order of magnitude, from several hundred thousand dollars to several ten thousands. Such systems would find uses in forensic science and the drug development industry.

Finally, there are also plenty of applications for deep UV light devices from around 280 to 254 nanometers. This is the solar-blind region, where light from the sun does not penetrate. Radiation below 290 nm destroys DNA. “So now you have a tool that is good for disinfection, decontamination, and sterilization of air and water with all the advantages of being a digital source,” Gaska explains. What are these advantages? One is that, because a mercury lamp requires time to warm up, it must always be switched on—even when you are not using it. An LED, by contrast, produces instant light because it turns on in less than a nanosecond. Having to keep mercury lamps always on means you have to replace them more often, because they burn out. LEDs are intrinsically long-life devices to begin with. They don't have to be kept on all the time, so they effectively last forever. In military jargon, LEDs are “fire and forget.”

Disinfecting water by destroying microbes with ultraviolet light has been used for many years, primarily at bottling plants. More recently, the technology has been adopted by some municipal water districts. LEDs would offer significant advantages for this application, such as dramatically lower running costs (electricity being the largest). However, it will be probably be some years before UV LEDs can match mercury lamps for the sheer output power needed to purify large amounts of water. Here, however, is an example of where the technology can radically alter the nature of the application. Why disinfect at the reservoir? Would it not be better to disinfect at the point of use, in the home? After all, no self-respecting terrorist is going to waste his time chucking anthrax into the municipal water system, where he knows it will get killed. It makes more sense to attack an unprotected part of the system, such as the water pipes leading from the supply. But stick a UV LED in a home faucet and it will zap anything nasty that comes in.

“You can have distributed networks,” Gaska predicts, “where you do your personalized water purification whenever you need it. Say you have a coffee maker. You'll have three UV LEDs installed there and they will do the job. You'll get your gallon of sterilized water in twenty minutes. Instead of going and buying gallons of water from the grocery store, you'll have a homebuilt system that you'll know for sure kills bugs. And this is a big problem, because with filtration you don't get rid of bugs, you remove particles and stuff, but not viruses or bacteria. Instead of a fifty-dollar mercury lamp, you can put in a tiny two-dollar LED, and you don't need much power. Once we get to the point when we can make these LEDs much cheaper and more efficient, we keep increasing the flow. Then, as the technology matures, we will go from a quarter gallon per minute, to half a gallon, to a gallon.” And on and on, to bigger and bigger applications. One company, Ohio-based Oh Technology, is already using SET's LEDs to treat raw sewage on an experimental basis. Early tests have proved viability, reducing bacterial levels by 60 percent.

“The applications are there,” Gaska insists. “For example, look at what's going on with the air and water supplies in airplanes. When you have 350 people packed into one place, and one person is sick, you're not putting mercury lamps up there, that's for sure. But LEDs? No hazard source, no high voltage, no mercury, no toxicity. And you're not talking about one or two planes—there's thousands of them. So we clearly see that UV LEDs are going to be really big. From the application standpoint, it will enable a lot of applications, and a lot of markets. At this point we are creating those markets, because we are providing people with new technology, and they're going through their design cycles. Now customers are coming back, and orders are gradually increasing. Already we are receiving orders for hundreds of thousands of devices. Once volume starts picking up, there are no fundamental reasons why the devices should be expensive. I think UV LEDs will be comparable to visible LEDs. We've showed everybody that it's feasible.” Other companies are now following the trail that SET has blazed. “We're a moving target,” Gaska concluded, in anticipation of the competition. “We will not be alone.”

Part 4 discusses the ongoing revolution in solid-state lighting more broadly. First, however, we must return to witness the dramatic second act in the life of the man who launched that revolution, Shuji Nakamura.